Those findings do not explain every decline. They show why your audit needs to connect search visibility with what real visitors do next. This guide walks you through a practical audit: how to find the pages that matter, price each loss in missed conversions, diagnose the real cause and measure recovery without fooling yourself.

Key numbers before you start

• 68% of US Google searches ended without a click in January to April 2026, up from 60% in 2024. (Source: SparkToro)
• 58% lower CTR for the page ranking first when an AI Overview appears. (Source: Ahrefs, 2026)
• 26% vs. 16%: users ended their browsing session more often on Google pages with an AI summary than on pages without one. (Source: Pew Research Center)
• 120% more clicks per impression for brands cited inside an AI Overview than for uncited brands on the same results page. (Source: Seer Interactive)

More data: our zero-click search statistics and AI search statistics roundups.

In this guide

• Identify the traffic that supports your business
• Build a recovery plan around conversion value
• Separate visibility loss from click loss
• Match your fix to the evidence
• Measure recovery without creating new problems
• A 60-minute traffic loss audit checklist
• FAQ

Identify the Traffic That Supports Your Business

A large traffic drop does not always deserve your first response. A glossary article might attract thousands of casual readers, while a smaller comparison page helps buyers choose your product. If new informational posts grow, they can hide a decline in pages that produce inquiries or sales.

For example, your blog could gain visitors overall while its software comparison guide loses qualified readers. Your headline traffic report would look positive, but your business would receive fewer opportunities.

You can define valuable traffic through a specific business action, such as a qualified inquiry, purchase, or newsletter signup. In your analytics platform, compare organic landing pages, the pages visitors first reach through unpaid search, against those actions. This gives your audit a business purpose instead of treating every click as equal.

A simple way to start is to sort your pages into three groups before you look at any decline:

Reach pages still matter, because they earn links and AI citations that lift the rest of the site. But they are also the pages most exposed to AI answers, so a fall there is often expected. Our guide to a zero-click SEO strategy explains how to measure those pages by visibility instead of sessions.

Build a Recovery Plan Around Conversion Value

Once you identify valuable pages, percentage drops can still distort your priorities. A page losing half its visits may matter less than another losing a smaller share. For example, a general marketing definition and a buying guide serve different readers, even when both cover the same subject.

SparkSeo is a search engine optimization agency whose seo services connect keyword research, content improvements, and technical optimization with conversions. That approach provides a useful principle for your recovery plan: assess the visitor’s route from search query to business outcome, rather than judging rankings alone.

You can regularly export page-level clicks from Google Search Console and organic landing-page sessions from your analytics platform. Compare the latest complete 28 days with the previous 28 days, then check the same season last year. Equal periods reduce weekday differences, while seasonal context helps prevent unnecessary changes.

For prioritization, multiply lost organic sessions by the page’s historical conversion rate. A hypothetical guide losing 500 sessions at a 4% rate represents 20 estimated missed conversions; a glossary losing 2,000 at 0.2% represents four. These estimates assume visitor quality remains similar, so they guide investigation rather than prove lost sales.

A simple scoring sheet

Put every declining page into one spreadsheet with these columns, then sort by the last one:

Pages that link into your money pages deserve a second look even if they convert poorly themselves. If an assist page loses traffic, fewer readers reach the page that sells. Check that with a landing page path or “next page” report.

Separate Visibility Loss from Click Loss

A page can lose clicks because fewer people search, its rankings fall, or its listing attracts less attention. These problems can produce similar totals but need different responses. For example, an article may hold its position while an AI summary answers the reader’s basic question before they visit.

Compare Queries Before Changing Content

Search Console shows impressions, the number of times your result appears, and click-through rate, the share of impressions producing clicks. You can filter a declining page and compare its main queries across matching countries and devices.

Falling impressions do not prove falling demand because weaker visibility also reduces impressions. You can compare several established pages covering the same topic and check seasonal patterns. This broader comparison helps you avoid rewriting a useful page because interest temporarily fell.

Google Trends is a quick second check: if interest in the main query fell by a similar share in the same months, the page is probably fine and the market moved.

Inspect the Results Your Readers See

Stable impressions with fewer clicks suggest a click-through problem, but average position can hide changes between queries. You can review individual searches and note whether competing results, advertisements, or AI summaries change the page layout.

A specific title can help readers recognize your page’s value. For example, a guide with tested email templates has a clearer reason to earn a visit than another generic outreach overview. The title should describe content you actually provide, because misleading promises weaken trust.

When an AI Overview sits above your result, the goal shifts to being one of the sources it cites. Seer Interactive found cited brands get 120% more clicks per impression than uncited ones. Our breakdown of how Google AI Overviews choose sources explains what tends to get a page cited, and our list of AI search monitoring tools shows how to track it.

Quick diagnosis table

Match Your Fix to the Evidence

A content refresh cannot repair every decline. Your guide might contain outdated pricing, miss the reader’s current question, or become inaccessible after a recent website update. A redesign could also remove useful internal links, links between pages on your website, making an important page harder for readers and search engines to find.

You can check whether Google still indexes the page, meaning it includes the page in its searchable collection, through Search Console’s URL Inspection tool. If the decline follows a release, your developer can check accidental indexing restrictions and broken redirects, which should send visitors to another address.

Your team can also check mobile rendering, or content display on phones, before rewriting. This order addresses access failures first.

For content problems, you can compare your examples with the questions readers now ask. Updated screenshots, clear limitations, and original experience give visitors practical reasons to stay. Health or financial content also needs qualified review because errors can affect important decisions.

Relevant internal links can restore useful routes from related articles. If you run WordPress, our guide to WordPress internal linking shows a quick way to add them. Backlinks, links from other websites, deserve a relevance check before further outreach; a backlink audit shows whether the page lost relevant backlinks or picked up toxic ones. More links cannot correct an inaccessible page or an article that answers the wrong question.

Fix in this order

1. Access: indexing, redirects, robots rules, canonical tags, mobile rendering.
2. Tracking: make sure the analytics tag and conversion events still fire on the page.
3. Intent: does the page still answer what people search for today? Compare it with the current top results.
4. Freshness: prices, screenshots, years, statistics and examples.
5. Snippet: title and meta description that state the page’s specific value.
6. Links: internal links from related pages first, then earned links from relevant sites.

If the decline is on product pages, our list of practical SEO tactics for product pages covers fixes specific to them. SaaS teams should also check the common SEO mistakes by SaaS startups, many of which show up as slow, quiet traffic loss.

Measure Recovery Without Creating New Problems

An early increase can tempt you to declare success, especially after several changes. However, seasonal demand or a tracking update can create movement that your edits did not cause. A page could also regain clicks while attracting fewer suitable buyers.

You can record each change, its date, and the intended outcome. A title change should be assessed through matched-query click-through rates; a content update also needs qualified conversions. A four-week review is a useful starting point, but low-volume pages need longer observation.

Make one kind of change at a time on a given page where you can. If you change the title, content and links in the same week, you will not know which one worked.

Keyword stuffing and irrelevant link campaigns introduce risk instead of solving the reader’s problem. You can protect quality by linking naturally and making every added paragraph useful. If you do build links to a recovering page, judge them by return beyond rankings, not by the number of links.

A 60-Minute Traffic Loss Audit Checklist

1. Export page-level clicks and impressions from Search Console: last 28 days, previous 28 days and the same period last year.
2. Export organic landing-page sessions and conversions from your analytics tool for the same periods.
3. Tag each page as a money, assist or reach page.
4. Calculate estimated missed conversions (sessions lost × historical conversion rate) and sort by it.
5. For the top five pages, check impressions, position and CTR by query to separate visibility loss from click loss.
6. Search the main queries yourself and note AI Overviews, ads, videos and forum results.
7. Run URL Inspection and check redirects, canonical tags and mobile rendering.
8. Pick one fix per page, log it with a date and a review date.
9. Review after four weeks (longer for low-traffic pages) using matched queries and qualified conversions.

FAQ

How do I find which pages are losing traffic in Google Search Console?

Open Performance, then Search results, choose Compare and select the last 28 days against the previous 28 days. Switch to the Pages tab and sort by the click difference. Repeat with the same period last year to rule out seasonality.

Why is my traffic dropping when my rankings are the same?

Usually because the results page changed. AI Overviews, ads, videos and forum results can push your listing down or answer the question before anyone clicks. Ahrefs measured a 58% lower click-through rate for the top result when an AI Overview appears.

Which lost pages should I fix first?

Fix the pages with the highest estimated missed conversions, not the biggest traffic drop. Multiply sessions lost by each page’s historical conversion rate and start at the top of that list.

How long does it take to recover lost organic traffic?

Review each change after about four weeks. Pages with low traffic need longer, often two to three months, before the data is reliable. Technical fixes such as broken redirects can show results faster.

Should I build more backlinks to a page that is losing traffic?

Only after you rule out access, tracking and intent problems. Links help a page that already answers the right question. They cannot fix a page Google cannot index or one that no longer matches what people search for.

Our best recovery work brings suitable people back to pages that help them decide. Higher traffic is valuable when relevance, transparency, and a better experience improve alongside it.

Related Reading

• Zero-click search statistics
• AI search statistics
• Zero-click SEO strategy
• SEO statistics
• Long-tail content

Amazon has never been more crowded. Big brands outbid everyone for the top sponsored slots, and many smaller brands with genuinely good products end up buried on page three. The right agency can change that. The wrong one can burn through your ad budget while sending you monthly reports.

We compared eight Amazon agencies serving brands on Amazon.com, Amazon.co.uk and the European marketplaces. We looked at who runs them, what they actually do, which markets they cover, the results they publish and who they suit best. Every figure below comes from the agency’s own website or case studies, checked in October 2026.

• The 8 best Amazon agencies at a glance
• How we compared them
• 1. Amazon Experts
• 2. Canopy Management
• 3. Sitruna
• 4. Lezzat
• 5. eStore Factory
• 6. Optimizon
• 7. Amazowl
• 8. Market Rocket
• How to choose the right Amazon agency
• FAQ

The 8 best Amazon agencies at a glance

How we compared them

• Hands-on Amazon experience: do the people running it sell on Amazon themselves, or only advise?
• Markets: can they run your brand on Amazon US, UK and Europe, or only one region?
• Scope: account management, advertising, listings and SEO, brand protection.
• Proof: do they publish real numbers from client accounts, with names?
• Who they suit: the size and type of brand they work best with.
• Value for money: what you get for your budget, and who actually does the work on your account.

1. Amazon Experts: best overall for value and hands-on seller experience

Best for: brands of any size that want people who sell on Amazon every day, not just people who advise on it.

Most agencies learn Amazon from the outside. Amazon Experts learned it from the inside. Founders Stanislav Yordanov and Veselin Angelov have been active Amazon sellers since 2018, with eight figures in Amazon sales and more than 50 brands in their own seller accounts. They cover Amazon UK, the US and six European marketplaces: Germany, France, Italy, Spain, the Netherlands and Sweden. Everything they offer clients is what they already do, every day, on their own accounts.

The Robin Hood idea. The agency sums up its mission in one image: Robin Hood taking traffic back from the giants. On Amazon, big brands with big budgets buy the front row, and good products from smaller makers get pushed out of sight. Amazon Experts sets out to win that visibility back through sharper listings, smarter ad spend and tighter control of who sells the brand, rather than simply outspending competitors.

Amazon Experts case studies: the numbers

Unlike most agencies on this list, Amazon Experts publishes named results with the numbers behind them:

Clients back this up in their own words. Paolo Conri, SEO Manager at Lacote Group, says their Amazon sales rose by more than 400% in three months after the team rebuilt their listings, SEO structure and advertising. Wojciech Szeliga, owner of Vis Plantis, says the account started “almost as a small test” and now brings in tens of thousands in monthly turnover.

Why Amazon Experts ranks first

• Sellers, not just an agency. They manage their own stock, ads, reviews and listings across eight marketplaces. Clients get advice tested on real accounts, with real money.
• Value for money. As a small, senior team they keep overheads low, so they can charge less than larger agencies without handing your account to a junior account manager.
• Founder-led quality. Three assistants handle routine work such as shipment plans, listing edits, SEO fixes and replenishment planning. That leaves both founders free to work on strategy, complex cases and growth for each client.
• Shared risk. In the distribution model they buy your stock and become the only authorised seller, so they only earn when it sells. You keep the listings, the content and Brand Registry.
• Attention to detail. On a top-selling colour conditioner, customers complained in reviews about stained hands. The fix was simple: gloves now come with every unit. It is the kind of detail sellers notice and outside consultants often miss.

Services: full-service 1P and 3P account management (Seller Central and Vendor Central), Amazon PPC and DSP, listing and SEO optimisation, A+ content and Brand Story, Amazon video, brand protection against unauthorised resellers, EU compliance, account audits, DTC/Shopify-to-Amazon launches and authorised distribution. They also offer free tools, including a one-ASIN listing audit, a title rewriter and a main image checker.

Pricing: not published. You get a written plan and a quote after a free 30-minute call.

Markets: Amazon UK, US, Germany, France, Italy, Spain, the Netherlands and Sweden.

Website: amazonexperts.eu

2. Canopy Management: best for US growth brands

Best for: growth-stage brands focused on Amazon.com.

Canopy Management is one of the best-known US Amazon agencies. Its homepage claims more than $3.2 billion in managed revenue and over 1,000 brands scaled. It offers account management together with advertising, inventory and account-health support, and it now also covers Walmart, Shopify, Google, Meta and TikTok.

Markets: mainly Amazon US. Website: canopymanagement.com

3. Sitruna: best for premium product pages

Best for: brands that want polished, high-end Amazon product pages alongside account management.

London-based Sitruna offers Amazon account management and advertising, with a particular focus on premium content and product page quality. Its homepage lists it as an Amazon Ads Advanced Partner, an Amazon Marketplace EU Agency Platinum Partner and part of Amazon’s Solution Provider Network. Clients named in agency directories include Antler, aden + anais and Lucky Saint.

Markets: mainly Amazon UK, plus the EU. Website: sitruna.com

4. Lezzat: best for PPC and Europe expansion

Best for: brands that need specialist Amazon PPC, including US sellers looking at Europe.

Lezzat, based in Reading with an office in Athens, specialises in Amazon PPC and advertising strategy. It is a verified Amazon Ads and SPN partner, works in Seller Central and Vendor Central, and says it has 7+ years of Amazon-only experience. It also helps US sellers expand to the UK and European marketplaces.

Markets: UK, Europe and US sellers expanding to Europe. Website: lezzat.co.uk

5. eStore Factory: best for small and mid-sized sellers

Best for: small and mid-sized sellers that want listings, PPC and creative under one roof.

eStore Factory covers listings, A+ content, PPC, storefront design and account-health support. It is an Amazon SPN partner and says it has helped more than 2,000 sellers, with a large in-house team that suits sellers who want many services at a lower price point.

Markets: mainly Amazon US. Website: estorefactory.com

6. Optimizon: best for profitable scaling in the UK

Best for: established brands that want to scale on Amazon while keeping a close eye on profitability.

UK-based Optimizon, with offices in London and Swindon, provides Amazon consultancy and marketplace management aimed at profitable growth, including advertising, DSP, listings and Brand Registry. Smollan bought Optimizon in October 2025, and in 2026 the agency began moving to the Smollan Digital Commerce brand. That is worth knowing if you prefer a large global group behind your agency.

Markets: mainly Amazon UK. Website: optimizon.co.uk

7. Amazowl: best for food, beverage and supplement brands

Best for: food, drink (including alcohol) and nutrition brands, and brands that need listings in several languages.

Amazowl now positions itself as an Amazon and retention agency for food, beverage and supplement brands in the US and UK. It also offers multilingual listing translation in German, Spanish, French, Italian, Portuguese, Swedish, Dutch and Polish, plus Amazon training.

Markets: US and UK, with multilingual European listings. Website: amazowl.com

8. Market Rocket: best for multi-channel marketing

Best for: brands that want Amazon handled together with Google, Meta and TikTok marketing.

London-based Market Rocket offers Amazon management and extends into Google, Meta and TikTok Shop for brands that want one partner across channels. It is an Amazon SPN partner and a TikTok Shop partner. The FMCG investor S-Ventures bought it for £2.3 million, and it still runs as a separate business that takes on outside clients.

Markets: mainly Amazon UK. Website: marketrocket.co.uk

How to choose the right Amazon agency

Before you sign with anyone, ask these five questions:

1. Do you sell on Amazon yourselves, and on which marketplaces?
2. Who will actually work on my account day to day: a founder, a senior specialist or a junior account manager?
3. How do you decide when to raise or cut ad spend, and how will I see the results (ACoS, TACoS, organic rank)?
4. What do you do about unauthorised resellers undercutting my price?
5. What does the first 90 days look like, and what do you need from me?

An agency that answers these clearly, with named examples and numbers from accounts it runs, is usually a safe bet. Be wary of anyone who only shows revenue growth without ad spend: a high sales number at a 50% TACoS can still lose you money.

If you want people who sell on Amazon themselves, offer good value and can run your brand in the US, UK and Europe from one team, Amazon Experts is our first pick. If you need a single-market specialist, specialist PPC or multi-channel marketing, the other agencies on this list are worth a call.

Amazon is only one channel. If you also sell on your own store, our Shopify statistics and eCommerce statistics for 2026 show where online shoppers are heading, and our guide to SEO tactics for product pages covers the Google side. Looking for help outside Amazon? See our lists of the best SEO agencies, niche SEO agencies for eCommerce and digital PR agencies.

FAQ

What is the best Amazon agency in 2026?

For most brands, Amazon Experts is our top pick: its founders have sold on Amazon since 2018, run 50+ brands in their own accounts and cover the UK, US and six EU marketplaces. Canopy Management is the biggest option for US-only brands.

How much does an Amazon agency cost?

Most agencies, including Amazon Experts, do not publish prices. Expect a monthly retainer, a percentage of sales or ad spend, or a mix of both. Some, like Amazon Experts, also offer a distribution model where they buy your stock and earn only when it sells.

What is a good ACoS for an Amazon agency to hit?

It depends on your margin and category. As a reference point, Amazon Experts published a 3.30% ACoS on €100,265 of ad sales for an Italian body-care brand in July and August 2026, and 16.83% for GUAM’s rebuilt UK campaigns. Always judge ACoS alongside TACoS and profit.

Which Amazon agency is best for selling in Europe?

Amazon Experts covers the UK and six EU marketplaces (Germany, France, Italy, Spain, the Netherlands and Sweden) from one team, including EU compliance. Lezzat is a good PPC specialist for US sellers expanding to Europe.

Should I hire an Amazon agency or build an in-house team?

An in-house hire is one person covering SEO, ads, design, stock and compliance. An agency gives you a whole team for a similar cost. Pick an agency whose people sell on Amazon themselves and work on your account directly.

Disclosure: we know the founders of Amazon Experts personally. All agencies were assessed on the same criteria, and every figure comes from the agency’s own website, case studies or public reports, checked in October 2026.

For brands, that makes Copilot the most familiar AI engine to work with. If Bing can crawl, index and trust your pages, they are eligible for Copilot. Brands get named when they appear clearly in the pages Bing retrieves: retailer listings, reviews, “best X” articles, reference sites and their own websites.

Copilot is also unusual in one way: Microsoft gives site owners citation data. Bing Webmaster Tools now shows how often your pages are cited in Copilot. Microsoft is still clear that nothing guarantees a citation. This guide covers what Microsoft has said, what the 2026 data shows and what a brand can do.

• The short answer: what decides a Copilot citation
• How Copilot finds and cites pages, step by step
• How citations appear in Copilot answers
• Measuring citations: the Bing AI Performance report
• Bingbot, robots.txt and the controls that affect Copilot
• Which sites Copilot cites most in 2026
• Listicles, reviews and “best X” articles
• Freshness: does Copilot prefer new content?
• Which brand signals matter for Copilot mentions
• What a brand can do: a 9-step plan
• Limits: what no one can guarantee
• FAQs
• Get your brand into the sources Copilot reads

The short answer: what decides a Copilot citation

Here is the whole process in one table. Each row is covered in detail below.

How Copilot finds and cites pages, step by step

1. Copilot sends grounding queries to Bing

Microsoft calls the process of fetching web evidence for an AI answer “grounding”. In an October 2025 post, Microsoft wrote that experiences like Microsoft Copilot are “powered by Bing’s search index”.

Grounding queries do not look like what users type. In a study of one site’s Copilot data (July to October 2026, 328,700 citations across 1,319 pages), the median grounding query was 4 words long. Only 8.4% of the top 154 queries were phrased as questions. 14.9% asked for evaluation, with words like “review”, “legit”, “best” or “compare”. Another analysis by Search Influence (February 2026) described the queries as machine-written keyword strings.

The lesson is that short, keyword-style rankings in Bing still matter, even though users ask Copilot long questions.

2. Grounding uses the same index as Bing search

In a May 2026 post on the Bing blog, Microsoft explained that grounding shares the same crawlers and quality signals as search. What changes is the unit of value. Instead of ranking whole pages, grounding looks for “groundable” information: facts with clear provenance that can support a specific statement in an answer.

Microsoft’s grounding is not only used by Copilot. Jordi Ribas wrote on the Bing blog in February 2026 that Microsoft grounding “powers nearly every major AI assistant in the market”. As we covered in our guide to how ChatGPT chooses sources, OpenAI names Microsoft among its search providers.

3. Copilot’s citations match Bing more closely than other assistants

An Ahrefs study of 15,000 long-tail prompts (August 2025) compared the URLs each AI assistant cited with Google’s and Bing’s top 10 results. Across assistants, the average overlap with Bing’s top 10 was about 10%. Copilot had the highest Bing overlap of any assistant tested.

That is still a minority of citations. Grounding queries differ from the user’s original question, so Copilot often cites pages that do not rank for the question itself.

4. Copilot breaks pages into pieces

Microsoft says Copilot parses content into smaller pieces, evaluates them, and assembles the best ones into an answer (Microsoft Advertising). Its advice follows from that:

• Use clear titles, H1s and meta descriptions.
• Structure pages with headings, Q&A sections, lists and tables.
• Add schema markup where it fits.
• Avoid long walls of text, key information hidden in tabs, information only in images, and key facts locked in PDFs.

5. Copilot checks, loops and sometimes abstains

Microsoft’s May 2026 post adds three details. Retrieval can run in loops, with follow-up searches when the first evidence is not enough. Conflicts between sources need to be detected. And the system can abstain when the evidence is too weak. Stale or contradictory facts on your pages work against you, because they make your content harder to use as evidence.

How citations appear in Copilot answers

Microsoft described the citation format when it introduced Copilot Search in Bing in April 2025:

• Sources are cited prominently, with a list of every link used to build the answer.
• Inline links can sit on whole sentences or passages, not just a small number.
• Cited sources appear at the top and bottom of the answer.

Copilot Search launched worldwide except in Russia and China. Citations appear in the Copilot app, in AI-generated summaries in Bing, and in some partner products that use Bing grounding.

Microsoft’s rewritten Bing Webmaster Guidelines also make a point worth repeating to stakeholders: a decline in clicks does not always mean lost visibility. Users may read your content in a Copilot answer without visiting your site.

Measuring citations: the Bing AI Performance report

Bing Webmaster Tools added an AI Performance report as a public preview on February 11, 2026. It shows how often your pages are cited across Microsoft Copilot, AI-generated summaries in Bing and selected partner integrations.

In June 2026, Microsoft added four views. Intents groups grounding queries by purpose, such as informational, commercial or local. Topics groups them by subject. Citation Share shows your share of all citations for a grounding query. A Compare view was added at the same time.

Know the limits. Microsoft says the report does not show placement or ranking inside an answer. Citation Share does not reveal which competitors take the rest, and it is not a share of traffic. The report has no click data.

What early users have found

• Citations are concentrated. Search Influence looked at 91 days of its own data: 19,717 citations across 86 pages. One page took 69% of citations, and the top four took 90%.
• Citations are volatile. In the same data, citations peaked at 5,804 on December 7, 2025, then fell by 97% by February 2026.
• Queries are short and varied. In Lilach Bullock’s single-site study, 23 of the top 154 grounding queries were in languages other than English, covering 13.4% of citations.
• Content changes are hard to read. The same study added matching sections for 15 grounding queries. 13 gained citation share, with a median gain of 5.2 points. But queries that were not touched also rose, by a median of 4 points. The author does not claim cause and effect.

These are single-site case studies, so treat them as examples of what the report can show, not as benchmarks.

Bingbot, robots.txt and the controls that affect Copilot

Copilot does not have its own named crawler for grounding. It relies on Bing’s index, so Bingbot is the bot that matters. The Bing Webmaster Guidelines, rewritten in February 2026 and covered by Search Engine Journal, explain how standard controls affect Copilot and grounding.

Other points from the guidelines:

• Do not block Bingbot from pages you want cited, and keep pages reachable through crawlable links.
• Use IndexNow and XML sitemaps with accurate lastmod dates, so Bing picks up changes quickly.
• Consolidate duplicate pages, so signals point to one URL.
• Local businesses should keep Bing Places listings accurate, as the AI Performance announcement recommends.

Many sites have never reviewed how they treat crawlers and AI bots. Our study of AI bots in SaaS robots.txt files found that 73% of top SaaS sites name no AI bot at all. You can check your own file with the free robots.txt AI bot checker on our tools page.

What about llms.txt? Microsoft’s guidelines do not mention it. Adoption is growing anyway: 74% of 47 leading B2B SaaS sites serve an llms.txt file. Treat it as optional housekeeping, not a ranking lever. IndexNow and clean Bing indexing are the proven basics for Copilot.

Which sites Copilot cites most in 2026

The freshest public snapshot is Ahrefs’ list of the 50 most-cited websites in Copilot, based on more than 3 million US queries across all topics in September 2026.

Three things stand out:

• Retailers lead. Amazon and Walmart take almost 30% of the top-50 share between them, with Target, Etsy and Best Buy in the top 15. For product categories, accurate retailer listings are a direct source of Copilot citations.
• Smaller sites can break in. Ahrefs notes several low-authority sites in the top 50, including one with a Domain Rating of 5. Copilot appears to care more about the passage than the domain.
• The list is consumer-heavy. It covers all topics, so B2B sources look smaller than they are in a software category. Business and review sources in the top 50 include Forbes, LinkedIn, Consumer Reports and Trustpilot.

The lesson: check which sources Copilot cites for your buyer prompts. For your own site, the AI Performance report shows which pages already get cited.

Listicles, reviews and “best X” articles

“Best X” articles already have the shape of the answer Copilot builds: a shortlist of named options with reasons. They also match how Copilot is told to read content. Microsoft recommends lists, tables, clear headings and Q&A sections, which is the standard format of a good comparison article.

What the data says

• Evaluation queries are common. In the Lilach Bullock study, 14.9% of the top grounding queries used words like “review”, “best” or “compare”.
• Review sites are in the top 50. Consumer Reports, Trustpilot and BBB all appear in Ahrefs’ Copilot list.
• Bing rankings carry over. Copilot’s citations overlap with Bing’s top 10 more than any other assistant’s in the Ahrefs study, so lists that rank in Bing have an advantage.

Which lists are worth being on

We found no Copilot-specific study of which listicles get cited. Based on Microsoft’s own guidance, these traits should help:

1. The list ranks in Bing for short “best [category]” and “[brand] review” style queries.
2. Each entry is easy to lift. A clear heading, a short description, key facts and a “best for” line.
3. The facts are current. Microsoft says stale facts lead to wrong answers and that conflicting sources are detected.
4. The facts are verifiable. Bing’s guidelines ask for content that can be checked on its own, with clear entities and key information near the top.

So the goal is not “get on any list”. It is to be included, with accurate detail, in independent, well-maintained comparisons that rank in Bing. That is what our brand mentions service focuses on.

Retailers, reference sites and review platforms

• Retailers lead Copilot’s citations for product queries. Keep titles, specifications and descriptions accurate on every marketplace you sell through.
• Wikipedia is third in the Ahrefs list. Most brands do not qualify for an article, and Wikipedia’s conflict-of-interest rules discourage editing your own page. Earn the press coverage that would make you notable instead.
• Review platforms such as Trustpilot and BBB appear in the top 50. Up-to-date profiles with recent reviews are basic hygiene.

Freshness: does Copilot prefer new content?

Microsoft’s own statements point to freshness as a quality issue rather than a ranking bonus. Its May 2026 post says stale facts produce wrong answers. Its guidelines recommend IndexNow and accurate lastmod dates so Bing learns about changes quickly.

We found no public study that measures the age of the pages Copilot cites. The case studies do show that citations move fast. Search Influence saw a 97% drop within about two months. Lilach Bullock saw daily citations grow from 539 to 11,225 over three months on another site.

In practice:

• Use IndexNow, so Bing is notified of updates instead of waiting for a re-crawl.
• Update pricing, features and comparison pages when facts change, and make sure the date reflects a real change.
• Remove or fix outdated facts that contradict your current pages. Conflicting sources make your content harder to use.

Which brand signals matter for Copilot mentions

Citations are about pages. Mentions are about brands. We found no large correlation study of brand signals for Copilot specifically. Microsoft’s guidance does give clues:

• Clear entities. Bing’s guidelines ask for content with clearly identified entities. Use your brand name, product names and category consistently.
• Provenance. Grounding looks for facts with clear sources. A brand described the same way across its own site, retailers, reviews and the press is easier to cite with confidence.
• No manipulation. The 2026 guidelines expand Bing’s abuse rules to cover prompt injection, AI manipulation, link schemes and scaled content published without oversight.

This is why digital PR and earned mentions sit at the centre of AI visibility work. Bing still uses crawlable links to discover pages, so link building supports the retrieval step. But the mention on a trusted page is what gets your brand named.

What a brand can do: a 9-step plan

1. Verify your site in Bing Webmaster Tools. Check indexing, crawl errors and the AI Performance report.
2. Allow Bingbot and check meta tags. Make sure key pages are not blocked, noindexed or marked noarchive by mistake.
3. Turn on IndexNow. Keep sitemap lastmod dates accurate.
4. Read your grounding queries. Look for short, keyword-style queries where you are cited, and where your pages should be cited but are not.
5. Structure key pages. Put key facts near the top, use headings, lists, tables and Q&A sections, and keep important information out of images, tabs and PDFs.
6. Build a prompt list. Write 20 to 50 prompts a buyer would ask Copilot, run them monthly, and note cited sources and named brands.
7. Fix your retailer and review profiles. Accurate marketplace listings and recent reviews feed the sources Copilot cites most.
8. Get into lists that rank in Bing. Pitch independent “best X” articles and reviews with accurate, useful information.
9. Earn press with original data. Data-driven content and digital PR create mentions on trusted sites that Bing indexes quickly.

Limits: what no one can guarantee

Be careful with anyone who promises Copilot placements.

• Microsoft says there is no guarantee. Its October 2025 guidance says there is “no secret sauce that guarantees selection”. Bing’s guidelines say the same about citations.
• The report is a sample. Grounding queries in Bing Webmaster Tools are a sample, and the report does not show ranking inside answers or clicks.
• Citations swing fast. One site lost 97% of its citations in about two months. Short windows can mislead.
• Answers vary. Copilot can run follow-up searches, abstain or pick different sources for the same question.
• Most evidence is correlational or from single sites. Even content changes that coincide with gains may not be the cause.

A sensible goal is to raise the share of relevant prompts where your brand appears and the citations shown in Bing Webmaster Tools, tracked over months.

FAQs

Does Copilot use Bing or Google?

Copilot uses Bing. Microsoft says experiences like Microsoft Copilot are powered by Bing’s search index, and grounding shares the same crawlers and quality signals as Bing search. If your pages are not indexed in Bing, they are very unlikely to be cited in Copilot, whatever your Google rankings.

Can I see how often Copilot cites my site?

Yes. The AI Performance report in Bing Webmaster Tools, launched in February 2026, shows total citations, cited pages, a sample of grounding queries and trends. Since June 2026 it also shows intents, topics and citation share. It does not show clicks or where you appear inside answers.

Can I stay in Bing but keep my content out of Copilot?

Yes. Bing’s guidelines say the noarchive tag stops content being used in Copilot answers and grounding while the page can stay in Bing search. The nocache tag limits Copilot to the URL, title and snippet. A noindex tag removes the page from Bing and Copilot entirely.

Does IndexNow help with Copilot?

IndexNow tells Bing straight away when a page is added or changed, so updates reach the index used for grounding faster. Microsoft recommends it in both its guidelines and the AI Performance announcement. It does not guarantee a citation. It removes delay between your update and Bing seeing it.

Why does Copilot cite Amazon and Walmart so often?

Ahrefs’ September 2026 data covers broad US queries, many of them about products, so retailers lead. Product pages hold specific, structured facts such as prices and specifications, which suit grounding. For consumer brands, accurate marketplace listings are part of Copilot visibility. B2B brands should check their own prompts.

Do I need different content for Copilot than for Google?

Mostly no. Bing’s guidelines say SEO fundamentals still apply. The differences are emphasis: get indexed in Bing, use IndexNow, put key facts near the top, and write sections that make sense alone. Copilot’s grounding queries are short, so keyword-style rankings in Bing still count.

Get your brand into the sources Copilot reads

Copilot names the brands it finds in the pages Bing trusts: lists, reviews, retailers and press coverage. Create & Grow places brands in relevant “best X” articles, reviews and press pieces, and runs digital PR campaigns that earn mentions on sites AI assistants cite. Brand mention plans start at €1,500 a month.

We cannot guarantee that Copilot will mention you, and nobody honestly can. We can show you which sources it cites for your category and get you into the ones that matter. Contact us to start with a short audit of your prompts.

This guide is part of our series on AI search sources, which also covers ChatGPT, Perplexity and Claude.

More guides in this series on how AI search picks sources: ChatGPT, Google AI Overviews, Gemini, Perplexity, Claude.

For brands, this makes Claude different from ChatGPT and Perplexity. Claude searches less often, so training data carries more weight. When it does search, ranking well in Brave’s results is the clearest lever, and the pages it cites lean heavily towards listicles, reviews and opinion content rather than forums.

Anthropic does not publish ranking factors, and nobody can promise a Claude mention. What you can do is make sure Claude can reach your pages, and that your brand appears in the kinds of pages it tends to cite. The 2025 and 2026 data below shows what those are.

• The short answer: what decides a Claude citation
• Training data vs live search: when Claude searches
• How Claude’s web search finds and cites pages
• How citations appear in Claude’s answers
• Anthropic’s crawlers, Brave and robots.txt
• Which sources Claude cites most
• Listicles, reviews and “best X” articles
• Freshness: does Claude prefer new content?
• Which brand signals matter for Claude mentions
• What a brand can do: a 9-step plan
• Limits: what no one can guarantee
• FAQs
• Get your brand into the sources Claude reads

The short answer: what decides a Claude citation

Here is the whole process in one table. Each row is covered in detail below.

Training data vs live search: when Claude searches

Claude can talk about your brand without searching at all. In that case it relies on what the model learned during training. There is no citation and no way to see which pages shaped the answer.

Live search is where citations come from. According to Anthropic’s help page on web search, Claude uses a search tool when a question would benefit from current information. In the newer Claude experience there is no on/off switch. Claude searches when it judges that it helps, and users can type “Search the web” to make it search. On Team and Enterprise plans, an owner has to enable web search first.

Claude searches less often than ChatGPT. In data from Profound presented at its Zero Click conference in June 2026 and reported by Search Engine Land, Claude searched on 36.6% of prompts, against roughly 90% for ChatGPT. The type of prompt made a big difference:

The practical split for brands:

Buying questions (“best project management tool for agencies”, “X vs Y”) are exactly the kind that trigger a search. That is why most of this guide is about the search path. But because Claude answers more questions from memory, long-term presence across the web matters more here than in ChatGPT.

How Claude’s web search finds and cites pages

1. Claude rewrites the prompt into search queries

When Claude searches, it writes its own queries. The Profound data reported by Search Engine Land in June 2026 found that 94% of Claude’s fan-out queries included the current year, against 17% for ChatGPT. Claude’s fan-out queries were also stable: the same query appeared about 65% of the time when a prompt was repeated.

For brands, that suggests a clear target. Pages that rank for “[category] [current year]” style searches are the ones Claude is most likely to retrieve.

2. Results come largely from Brave Search

Anthropic has not published how its search works. The evidence that Brave Search supplies most results comes from three directions:

• The subprocessor list. In March 2025, Simon Willison noticed that Anthropic had added Brave Search to its list of subprocessors as of March 19, 2025, the day before web search launched. A test search he ran returned ten citations that exactly matched Brave’s results.
• Early overlap tests. Profound tested three “best X” queries in March 2025 and found that 13 of Claude’s 15 citations (86.7%) matched Brave’s top non-sponsored results (Profound). That was a very small sample.
• A larger 2026 test. Profound repeated the test in June 2026 on about 400 queries and found 79.2% of Claude’s citations came from Brave’s top 10 (Search Engine Land). Notes from the same session suggest Claude takes Brave’s top results with little re-ranking, though others dispute that.

There is one open question. An analysis by Xponent21 (July 2026) reports that Brave was still on Anthropic’s subprocessor list, and that Anthropic’s Trust Center added turbopuffer, a search infrastructure company, as a second web search subprocessor on May 6, 2026. Its role has not been made public. Brave remains the best-supported explanation for where Claude’s results come from.

3. Google rankings still correlate

Brave and Google often agree on the best pages. In the same Profound data, 64% of the sites Claude cited also appeared in Google’s top 50 for the same query, against 37% for ChatGPT. Strong Google rankings help, but checking your position in Brave is the more direct test.

4. Claude reads, filters and cites

Claude does not cite everything it retrieves. On Anthropic’s developer platform, the web search and fetch tools now write and run code to filter results before they reach the model. Anthropic says this dynamic filtering, released with Claude Opus 4.6 and Sonnet 4.6 in February 2026, improved accuracy by 11% on average while using 24% fewer input tokens on its search benchmarks.

Anthropic has not said how the consumer Claude app decides which retrieved pages to cite. The safe assumption is the same as for other engines: pages that answer the query directly and clearly are more likely to be used.

5. Claude and ChatGPT cite different pages

Do not assume that being cited in ChatGPT means being cited in Claude. Profound found Claude and ChatGPT shared only about 8% of citations in its 2026 data. Each engine needs its own check.

How citations appear in Claude’s answers

According to Anthropic, answers that use web search include citations so users can check the sources. These come as direct citations, links to the sources and relevant quotes.

A few details matter for brands:

• Location affects results. Claude infers a user’s approximate location from their IP address to localise results.
• Users can fetch your page directly. If a user pastes a URL, Claude’s web fetch retrieves that page. This is how Claude reads a product page someone is already looking at.
• Image search uses Bing. Anthropic says image search in Claude is powered by Bing, even though text search results match Brave.

Web search launched in March 2025 as a preview for paid US users and became available on all plans worldwide on May 27, 2025.

Anthropic’s crawlers, Brave and robots.txt

This is the part you control. Anthropic lists three bots in its crawler help page.

Key facts from Anthropic’s documentation:

• The three settings are independent. You can block ClaudeBot for training and still allow Claude-SearchBot and Claude-User.
• Anthropic says blocking Claude-SearchBot “may reduce your site’s visibility and accuracy in user search results”. Blocking Claude-User “may reduce your site’s visibility for user-directed web search”.
• The bots support the non-standard Crawl-delay rule, and Anthropic says they do not bypass CAPTCHAs or other anti-circumvention measures.
• Anthropic says blocking by IP address may not work reliably. Use robots.txt, set for each subdomain. Anthropic publishes its bots’ IP ranges for verification.

Brave’s crawler

Because Claude’s results match Brave so closely, Brave’s crawler matters too. According to Brave’s help page, its crawler does not use a separate user agent. If a page is not crawlable by Googlebot, Brave will not crawl it either. Brave also says robots.txt alone does not stop a page from being indexed; a noindex tag does. Brave offers a form to request a re-crawl of a URL.

In short: if Googlebot can reach your pages, Brave usually can. If you block Googlebot from a section, assume Claude will not see it through Brave.

Many sites have never made a decision about AI bots. Our study of AI bots in SaaS robots.txt files found that 73% of top SaaS sites name no AI bot at all. You can check your own file with the free robots.txt AI bot checker on our tools page.

What about llms.txt? Anthropic’s crawler page does not mention it as a ranking signal. Adoption is growing anyway: 74% of 47 leading B2B SaaS sites serve an llms.txt file. Treat it as optional housekeeping, not a ranking lever.

Which sources Claude cites most

There is no public top-50 list of Claude’s most-cited domains like the ones Ahrefs publishes for other assistants. What exists is a set of studies on the types of sources Claude prefers. They point the same way.

Three things stand out:

• Claude barely cites forums. Reddit dominates citation lists for ChatGPT and Perplexity. In the Profound data, community sites made up less than 1% of Claude’s citations.
• Reviews matter a lot. In Yext’s data, the share of Claude citations from reviews and social sources ranged from 6.3% for organisations to 24.4% for food and beverage. In food and beverage, that was about one in four Claude citations, against about one in forty for Gemini. In business services it was 15.9%, more than double any other model.
• Hard news is less central. Muck Rack found Claude cited Reuters 20 times less often than Gemini and 50 times less often than ChatGPT. The outlets it did cite lean towards buying guides and business advice.

The lesson: for Claude, review sites and well-ranked comparison articles in your category are more important than forum threads or breaking news.

Listicles, reviews and “best X” articles

“Best X” articles are the most direct route to a brand mention in any AI engine. In Claude, the data says they are even more central. Listicles made up 36% of Claude’s citations in Profound’s 2026 data, and “best X” prompts with a time element triggered a search 81% of the time.

Which lists are most likely to be cited

We found no Claude-specific study of which listicles get cited and which do not. Based on how Claude’s search works, these traits should help:

1. The list ranks in Brave. About four in five Claude citations came from Brave’s top 10 in Profound’s test. A list that does not rank there is unlikely to be read.
2. The title carries the current year. Claude adds the year to almost all of its search queries, so “best X in 2026” lists match its searches.
3. The list is maintained. A dated title with outdated content is a weak source, and it will drop as soon as the year changes.
4. The list gives reasons. Claude cited opinion content almost twice as often as ChatGPT. Clear verdicts and “best for” lines give it something to quote.

So the goal is not “get on any list”. It is to be included, with accurate detail, in independent, current comparisons that rank for the searches Claude runs. That is what our brand mentions service focuses on.

Review sites

Yext’s data makes review platforms a priority for Claude, especially for local and service businesses. Keep profiles on Google Business Profile, Yelp, Trustpilot, G2 or the sites that matter in your category up to date, with recent reviews and accurate details.

Freshness: does Claude prefer new content?

The signals are mixed.

• Its searches are dated. 94% of Claude’s fan-out queries included the current year in the Profound data. That pushes results towards pages written or updated for this year.
• Its journalism citations are older. Muck Rack found that 36% of Claude’s citations to news articles were from the past 12 months, against 56% for ChatGPT.

One way to read this: Claude favours current buying guides and comparisons, while news coverage counts more for the brand’s long-term profile than for any single answer. We found no study that measures the age of all pages Claude cites.

In practice:

• Update comparison and pricing pages when facts change, not just the date.
• Make sure the lists that mention you are refreshed for the current year.
• Do not expect a new page to beat an established, well-ranked page just because it is new.

Which brand signals matter for Claude mentions

Citations are about pages. Mentions are about brands. We found no correlation study of brand signals for Claude specifically. The largest such study, by Ahrefs on 75,000 brands, covers ChatGPT and Google’s AI features; we summarise it in our guide to how ChatGPT chooses sources.

Two things make brand signals especially important for Claude:

• It answers more from memory. Claude searched on about a third of prompts in Profound’s data. For the rest, what it knows about your brand comes from training. A brand that is widely and consistently described across the web has a better chance of being remembered.
• It reads third-party verdicts. Listicles, reviews and opinion pieces make up a large share of what it cites when it does search.

This is why digital PR and earned mentions sit at the centre of AI visibility work. Links help the pages that mention you rank in Brave and Google, so link building supports the retrieval step. But the mention itself is what gets your brand named.

What a brand can do: a 9-step plan

1. Allow Claude-SearchBot and Claude-User. Check robots.txt for each subdomain. Decide separately whether to allow ClaudeBot for training.
2. Do not block Googlebot from key pages. Brave follows Googlebot’s rules, so a Googlebot block also hides pages from Brave.
3. Check your Brave rankings. Search your main buyer queries on Brave Search, with the current year added. This is the closest public view of what Claude retrieves.
4. Build a prompt list. Write 20 to 50 prompts a buyer would ask, focused on “best X”, rankings and comparisons, since those trigger searches most often.
5. Record who gets cited. Run the prompts in Claude, note the cited sources and named brands, and repeat monthly.
6. Get into the lists Claude already cites. Pitch the independent “best X [year]” articles from step 5 with accurate, useful information.
7. Strengthen review profiles. Collect recent reviews on the platforms that matter in your category. Claude leans on reviews more than other models.
8. Earn press and expert mentions. Original data gives writers a reason to name you. Data-driven content and digital PR build the long-term presence Claude draws on when it answers from memory.
9. Keep your story consistent. Use the same name, category and positioning on your site, review profiles, LinkedIn and directories.

Limits: what no one can guarantee

Be careful with anyone who promises Claude placements.

• Anthropic has not published how search ranking works. The Brave link rests on its subprocessor list and on outside testing, not on an Anthropic statement about ranking.
• The setup can change. A second web search subprocessor was reportedly added in May 2026. Results could shift without notice.
• Answers vary. Whether Claude searches at all depends on the prompt. Location and model also change the results.
• Training data is slow. Work that changes how the model itself describes you only shows up after future model updates.
• Most studies are small or correlational. The first Brave overlap test used 15 citations. The later ones are larger but come from a single vendor.

A sensible goal is to raise the share of relevant prompts where your brand appears, track it over months and compare it with competitors.

FAQs

Does Claude use Brave Search?

Anthropic lists Brave Search as a subprocessor, and independent tests show strong overlap. Profound found 79.2% of Claude’s citations came from Brave’s top 10 in a June 2026 test of about 400 queries. Anthropic has not explained its ranking, and a second web search subprocessor was reportedly added in May 2026.

If I block ClaudeBot, will Claude stop mentioning my brand?

No. ClaudeBot collects content that could be used for training. Claude-SearchBot and Claude-User control search and user-requested visits. You can block ClaudeBot and still allow the other two. Claude can also name your brand based on other sites that mention you, or on earlier training data.

Why doesn’t Claude search for every question?

Claude decides when current information would help. In Profound’s 2026 data, it searched on 36.6% of prompts. “Best X”, ranking and comparison prompts triggered searches most often. “What is” and “how does” questions rarely did, so Claude answered those from what it learned in training.

How is Claude different from ChatGPT for brand visibility?

The two cite very different pages. Profound found they shared only about 8% of citations. Claude barely cites Reddit and other forums, cites listicles and opinion content more often, and adds the current year to most searches. Results line up with Brave Search rather than the providers ChatGPT uses.

Does Claude cite Reddit?

Rarely. In Profound’s 2026 data, forums and community sites such as Reddit and Quora made up less than 1% of Claude’s citations, against about 16% for ChatGPT. For Claude, review sites and well-ranked comparison articles are a better use of effort than forum activity.

How can I track whether Claude mentions my brand?

Run a fixed list of buyer prompts every month, focused on “best X” and comparison questions, and log the cited sources and named brands. Check the same queries on Brave Search. Also watch referral traffic from claude.ai in your analytics, since search answers link to their sources.

Get your brand into the sources Claude reads

Claude names the brands it finds in current lists, reviews and opinion pieces, and the brands it already knows from training. Create & Grow places brands in relevant “best X” articles, reviews and press pieces, and runs digital PR campaigns that earn mentions on sites AI assistants cite. Brand mention plans start at €1,500 a month.

We cannot guarantee that Claude will mention you, and nobody honestly can. We can show you which sources it cites for your category and get you into the ones that matter. Contact us to start with a short audit of your prompts.

This guide is part of our series on AI search sources, which also covers ChatGPT, Perplexity and Microsoft Copilot.

More guides in this series on how AI search picks sources: ChatGPT, Google AI Overviews, Gemini, Perplexity, Copilot.

For brands, that makes Perplexity one of the more predictable AI engines to work with. A page has to be open to PerplexityBot, sit in Perplexity’s index, and contain a passage that answers the question better than competing pages. Brands get named when they keep appearing in the pages Perplexity trusts: review sites, “best X” lists, YouTube, Reddit, Wikipedia and their own websites.

Nobody can promise a Perplexity citation. Perplexity does not publish its ranking factors, and answers change from run to run. What you can do is make your pages and your brand easy to find in the places it reads. The 2025 and 2026 data below shows where those places are.

• The short answer: what decides a Perplexity citation
• How Perplexity finds and ranks sources, step by step
• How citations appear in Perplexity answers
• Perplexity’s crawlers and robots.txt
• Which sites Perplexity cites most in 2026
• Listicles, reviews and “best X” articles
• Freshness: does Perplexity prefer new content?
• Which brand signals matter for Perplexity mentions
• What a brand can do: an 8-step plan
• Limits: what no one can guarantee
• FAQs
• Get your brand into the sources Perplexity reads

The short answer: what decides a Perplexity citation

Here is the whole process in one table. Each row is covered in detail below.

How Perplexity finds and ranks sources, step by step

Perplexity has described its search stack in unusual detail. In a September 2025 research article on its search API, it said the same infrastructure serves its own products, which handle 200 million daily queries. Its index tracks more than 200 billion unique URLs.

1. Perplexity uses its own index

The earliest version of Perplexity, in late 2022, relied on third-party search APIs. Today it runs its own crawler and index. Perplexity says machine learning models predict whether a URL should be indexed and when it should be crawled again, based on how important the page is and how often it changes.

Perplexity also says it prioritises keeping documents from authoritative domains “hot” in the index. In plain terms, well-known sites get crawled and refreshed more often than unknown ones.

This matters because Perplexity’s results differ from Google’s. In an Ahrefs study of 15,000 long-tail prompts (August 2025), 28.6% of the URLs Perplexity cited ranked in Google’s top 10 for the same query. That was the highest share of any assistant tested; the others were around 8%. So Google rankings help, but most Perplexity citations still come from outside Google’s top 10.

2. Pages are split into passages

Perplexity does not only rank whole pages. It splits documents into smaller sections and scores each one on its own. A strong paragraph on a long page can be retrieved even if the rest of the page is about something else.

This is the main practical lesson for content. Each section of a page should make sense on its own, with the key fact, number or recommendation stated plainly near the top of that section.

3. Retrieval mixes keywords and meaning

Perplexity’s research article describes a “hybrid” retrieval step. It runs a classic keyword search and a meaning-based (semantic) search together, then merges the results. Before ranking, a filter removes results that are stale or do not respond to the query.

Pages that use the exact words buyers use, and that also explain the topic in depth, have two ways to be found.

4. Ranking narrows the shortlist

Ranking happens in stages. Early stages use fast scoring based on keywords and embeddings. The final stage uses more precise models called cross-encoder rerankers, which compare the query and each passage closely. Perplexity says ranking also learns from signals in millions of user requests per hour.

5. Deeper modes read far more sources

Perplexity has several search modes. According to Perplexity’s help center (updated August 2026), Pro Search researches across multiple sources and lets subscribers choose the model, such as GPT-5 or Claude 4.6 Sonnet. Its Research mode “performs dozens of searches automatically, reads hundreds of sources” and returns a report in 2 to 4 minutes.

The model can change the wording of the answer. The sources still come from Perplexity’s search system, so the steps above apply whichever model a user picks.

How citations appear in Perplexity answers

Citations are the core of Perplexity’s product. Its help center says each answer “includes numbered citations linking to the original sources”, so users can check the information or read further (Perplexity).

Three things follow for brands:

• A citation and a mention are different. Your site can be cited without your brand being named, for example when your blog post explains a general topic. Your brand can also be named because a review site mentioned it, with the citation going to that review site.
• Citations are visible. Users see the source list for every answer. A cited review or listicle that describes you badly is shown to everyone who asks.
• Follow-up questions run new searches. Perplexity keeps the context of earlier questions, so a buyer who asks “best CRM for startups” and then “which one is cheapest” can get a new set of sources.

Perplexity’s crawlers and robots.txt

Perplexity lists two user agents in its crawler documentation.

Key facts from Perplexity’s documentation:

• Perplexity recommends allowing PerplexityBot in robots.txt and allowing its published IP ranges in your firewall or CDN. The documentation includes setup steps for Cloudflare and AWS WAF.
• Each setting works on its own. Changes can take up to 24 hours to take effect.
• Perplexity does not run a separate training crawler, so there is no “allow search, block training” choice like the one OpenAI offers.

There is one dispute to know about. In August 2025, Cloudflare reported that when test sites blocked Perplexity’s declared bots, it saw an undeclared crawler posing as Chrome on macOS fetching the content. Cloudflare counted 3 to 6 million such requests a day, against 20 to 25 million for the declared Perplexity-User agent, and removed Perplexity from its list of verified bots. This is Cloudflare’s account. For brands that want to be cited, the lesson is simpler: if you want Perplexity traffic, allow its bots openly rather than half-blocking them.

Many sites have never made a decision about AI bots at all. Our study of AI bots in SaaS robots.txt files found that 73% of top SaaS sites name no AI bot. You can check your own file with the free robots.txt AI bot checker on our tools page.

What about llms.txt? Perplexity’s crawler documentation does not mention it, although Perplexity’s own help center publishes one. Adoption is growing: 74% of 47 leading B2B SaaS sites serve an llms.txt file. Treat it as optional housekeeping, not a ranking lever.

Which sites Perplexity cites most in 2026

The most-cited domains show where Perplexity looks when it builds an answer. The freshest public snapshot is Ahrefs’ list of the 50 most-cited websites in Perplexity, based on more than 3.1 million US queries across all topics in September 2026.

Three things stand out:

• YouTube is almost level with Reddit. In ChatGPT, YouTube sits much lower. In Perplexity it takes about a fifth of the top-50 share. Video content with clear titles and descriptions is a real route into Perplexity answers.
• Review and buying-guide sites are well represented. Good Housekeeping, Consumer Reports, TechRadar, NerdWallet, CNET, PCMag and Tom’s Guide all make the top 50.
• B2B sources look small in a general list. In Profound’s analysis of 680 million citations (August 2024 to June 2025), Perplexity’s most-cited domains included Gartner (1.0%), LinkedIn (0.8%), Forbes (0.7%) and G2 (0.6%). Reddit led at 6.6%, almost half of the top-10 share on its own.

Brand websites matter too. Yext analysed 17.2 million citations from four AI models in Q4 2025, mostly for US location-based searches. It found Perplexity was the most consistent model across industries, with first-party websites taking 37% to 50% of citations in most sectors. Yext stresses that its data shows correlation, not cause.

The lesson: a general top-50 list is a starting point. Check which sources Perplexity cites for your buyer prompts, and target those.

Listicles, reviews and “best X” articles

“Best X” articles are the most direct route to a brand mention, because they already look like the answer Perplexity is building: a shortlist of named options with reasons. Perplexity’s passage-level ranking also suits them. A list entry with a clear heading, a short description and a “best for” line is exactly the kind of passage it can lift.

What the data says

• Buying-guide publishers are among the top sources. Ahrefs’ September 2026 list includes at least seven review and buying-guide sites in Perplexity’s top 50.
• Review platforms show up for B2B. G2, Gartner and PCMag are all in Profound’s Perplexity top 10.
• Google rankings carry over more than elsewhere. Perplexity’s 28.6% overlap with Google’s top 10 in the Ahrefs study means lists that rank well in Google have a better chance here than in other assistants.

Which lists are worth being on

We found no Perplexity-specific study of which listicles get cited. Based on how Perplexity says it ranks content, these traits should help:

1. Each entry stands alone. Perplexity scores passages, so each product needs its own clear section with the brand name, what it does and who it is for.
2. The list is maintained. Perplexity’s filters drop stale content and refresh important pages more often.
3. The publisher is established. Perplexity says it keeps documents from authoritative domains fresh in its index.
4. The list ranks for the right queries. A list nobody can find in search is unlikely to be retrieved.

So the goal is not “get on any list”. It is to be included, with accurate detail, in independent, well-maintained comparisons that rank for the questions your buyers ask. That is what our brand mentions service focuses on.

Reddit, YouTube and review sites

• Reddit leads both the Ahrefs and Profound lists. Check the threads in your category to see which brands users recommend and how they describe you.
• YouTube is a near-equal second in Ahrefs’ 2026 data. Reviews, comparisons and tutorials that name your brand in the title and description give Perplexity something to cite.
• Review sites such as G2, Trustpilot and Consumer Reports feed citations directly. Keep profiles current and collect recent reviews.

Freshness: does Perplexity prefer new content?

Perplexity’s own description points to yes, with a nuance. According to its research article, a filter removes stale content before ranking. Its indexing models also decide how often to re-crawl a page based on how often it changes. The system runs tens of thousands of indexing operations per second to keep the index current.

That is not the same as “newest wins”. Freshness affects whether a page is in the index and up to date. Relevance and passage quality still decide the ranking. We found no public study that measures the age of Perplexity’s cited pages, so be careful with anyone quoting a precise figure.

In practice:

• Update pricing, feature and comparison pages when the facts change, not just the date.
• Keep the pages you want cited linked from your homepage and sitemap, so PerplexityBot finds changes quickly.
• Expect freshness to matter most for news, pricing and “best X in 2026” questions.

Which brand signals matter for Perplexity mentions

Citations are about pages. Mentions are about brands. We found no large correlation study that looks at brand signals for Perplexity specifically. The best-known study, by Ahrefs on 75,000 brands, covers ChatGPT and Google’s AI features; we summarise it in our guide to how ChatGPT chooses sources. Its main finding was that branded web mentions and YouTube mentions correlate with AI visibility far more than links or page counts do.

The Perplexity data available points the same way:

• Perplexity reads widely talked-about sources. Reddit, YouTube, Wikipedia and Facebook make up more than half of the top-50 share in Ahrefs’ 2026 list.
• Your own site counts. Yext found first-party websites took 37% to 50% of Perplexity citations in most industries it studied.
• Authority affects crawling. Perplexity says it prioritises documents from authoritative domains when keeping its index fresh.

This is why digital PR and earned mentions sit at the centre of AI visibility work. Links still help your pages get discovered, crawled and ranked, so link building supports the retrieval step. The brand mention on a trusted third-party page is what gets you named.

What a brand can do: an 8-step plan

1. Allow PerplexityBot. Check robots.txt, then check that your CDN or firewall allows Perplexity’s published IP ranges.
2. Build a prompt list. Write 20 to 50 prompts a buyer would ask: “best [category] for [use case]”, “[competitor] alternatives”, “[your brand] vs [competitor]”, “[your brand] pricing”.
3. Record who gets cited. Run the prompts in Perplexity, note the numbered sources and the brands named, and repeat monthly.
4. Write in self-contained sections. Give every product, pricing and comparison section a clear heading and a direct first sentence. Perplexity ranks passages, not just pages.
5. Rank in Google too. Perplexity’s results overlap with Google’s top 10 more than any other assistant in the Ahrefs study. Classic SEO still feeds this engine.
6. Get into the lists and reviews Perplexity already cites. Pitch the independent “best X” articles and review sites from step 3 with accurate, useful information.
7. Publish on YouTube. Reviews, demos and comparisons with your brand in the title give Perplexity a source type it cites heavily.
8. Earn press with original data. Data-driven content and digital PR campaigns create mentions on sites Perplexity trusts and refreshes often.

Limits: what no one can guarantee

Be careful with anyone who promises Perplexity placements.

• Perplexity does not publish its ranking factors. Its research article describes the architecture, not the weights.
• Answers vary. The mode, the model, follow-up questions and the user’s context all change the searches Perplexity runs and the sources it cites.
• Study samples differ. Ahrefs used US queries across all topics. Profound covered a period ending in June 2025. Yext focused on location-based searches. None of them matches your category exactly.
• Most studies are correlational. They show what cited pages have in common, not what causes a citation.

A sensible goal is to raise the share of relevant prompts where your brand appears, track it over months and compare it with competitors.

FAQs

Does Perplexity use Google to find sources?

No. Perplexity used third-party search APIs in its first version in late 2022, but now runs its own crawler, PerplexityBot, and its own index of more than 200 billion URLs. Its results still overlap with Google’s top 10 more than other assistants do, so Google rankings remain a useful signal.

If I block PerplexityBot, will Perplexity stop citing my site?

Blocking PerplexityBot keeps your pages out of Perplexity’s index, so they are unlikely to be cited in normal answers. Perplexity-User may still fetch a page when a user asks about it directly, since Perplexity says it generally ignores robots.txt for those fetches. Your brand can also appear through other sites.

Does Perplexity use my content to train AI models?

Perplexity says PerplexityBot is not used to crawl content for AI foundation models. It surfaces and links websites in search results. Perplexity-User is not used for crawling or training either. Unlike OpenAI and Anthropic, Perplexity does not list a separate training crawler in its documentation.

Why does Perplexity cite YouTube so often?

Ahrefs’ September 2026 data shows YouTube with 20.8% of the top-50 citation share, close behind Reddit. Perplexity has not explained why. Video pages often have descriptive titles and descriptions that match buyer questions. For brands, reviews and comparisons on YouTube are a practical way into Perplexity answers.

Does the model I choose in Perplexity change the sources?

The model writes the answer, while Perplexity’s search system finds and ranks the sources. Different models can phrase answers differently and lean on different results. Research mode runs dozens of searches and reads hundreds of sources, so it usually cites more pages than a quick answer.

How can I track whether Perplexity mentions my brand?

Run a fixed list of buyer prompts every month and log the numbered sources and named brands. AI visibility tools from Ahrefs, Profound and others automate this at scale. Also check referral traffic from perplexity.ai in your analytics, since every Perplexity answer links to its sources.

Get your brand into the sources Perplexity reads

Perplexity names the brands it keeps finding in trusted lists, reviews, videos and press coverage. Create & Grow places brands in relevant “best X” articles, reviews and press pieces, and runs digital PR campaigns that earn mentions on sites AI assistants cite. Brand mention plans start at €1,500 a month.

We cannot guarantee that Perplexity will mention you, and nobody honestly can. We can show you which sources it cites for your category and get you into the ones that matter. Contact us to start with a short audit of your prompts.

This guide is part of our series on AI search sources, which also covers ChatGPT, Claude and Microsoft Copilot.

More guides in this series on how AI search picks sources: ChatGPT, Google AI Overviews, Gemini, Claude, Copilot.

For brands, grounding is the part you can influence. Your pages need to be in Google’s index, rank for the searches Gemini runs, and not be blocked through the Google-Extended robots.txt token. Brands then get named when they appear often across the pages Gemini reads. In 2026 those were mainly Reddit, YouTube, Wikipedia, product review sites and big retailers.

Gemini is also more selective than Google’s other AI surfaces. It cites fewer sources per answer, and in one 2026 study it stopped citing sources in about a quarter of answers. Nobody can promise you a mention. The data below shows what raises the odds.

• The short answer: what decides a Gemini citation
• Training data vs grounding: two ways Gemini knows your brand
• How Gemini grounding finds and cites pages, step by step
• How sources appear in Gemini answers
• Google-Extended, Googlebot and robots.txt
• Which sites Gemini cites most in 2026
• Gemini vs AI Overviews and AI Mode
• Listicles, reviews and “best X” articles
• Freshness: does Gemini prefer new content?
• Which brand signals matter for Gemini
• What a brand can do: a 9-step plan
• Limits: what no one can guarantee
• FAQs
• Get your brand into the sources Gemini reads

The short answer: what decides a Gemini citation

Here is the whole process in one table. Each row is covered in detail below.

Training data vs grounding: two ways Gemini knows your brand

Like ChatGPT, Gemini has two sources of knowledge. The first is its training data, a snapshot of the web and other material from before the model was built. Anything Gemini says from training comes without a link. The second is live Google Search, used through grounding. Grounded answers come with sources you can click.

The split matters more than it seems. Seer Interactive tracked 82,000 Gemini responses across 20 brands from November 2025. Between mid-February and early March 2026, the share of responses with citations fell from 99% to 76%. Seer says the uncited answers rely on training data.

That means two kinds of work. Work on your training-data footprint, such as press coverage, Wikipedia and long-standing reviews, shapes what Gemini says without searching. Work on search visibility shapes what it cites when it does search. The first is slow and only shows up in future models. The second can move within weeks.

How Gemini grounding finds and cites pages, step by step

Google documents the grounding flow for developers in the Gemini API. The Gemini app is a separate product, but the API documentation is the clearest public description of how grounding works.

1. Gemini decides whether to search

The model reads the prompt and decides if a Google Search can improve the answer. Simple or timeless questions may not trigger a search. In Seer’s data, “how to” prompts kept citations in 100% of cases, while “best” and “top” recommendation prompts shifted most toward uncited answers.

2. Gemini writes its own search queries

If it decides to search, the model generates one or more queries automatically. These are often more specific than the user’s prompt. A question about “the best CRM for a small agency” may become separate searches for pricing, agency features and reviews.

3. Gemini reads the results

The model processes the search results and writes its answer. Developers can combine grounding with a URL context tool, which reads specific pages they supply.

4. Gemini links the supporting sources

The response comes back with citations that tie parts of the text to the pages behind them. Profound’s 2026 analysis found Gemini cited an average of 6.6 sources per run and named 4.4 brands per answer.

Gemini also has a large audience. In the Q2 2026 earnings remarks, Sundar Pichai said the Gemini app had 950 million monthly active users, and that daily users had tripled over the past year.

How sources appear in Gemini answers

Google’s Gemini Apps Help explains how sources show up:

• A Sources button appears at the bottom of a response or inline. It opens a side panel with the links.
• Sources can be public websites, files the user uploaded, or the user’s Google Workspace content.
• Not every response includes sources.
• If Gemini quotes a large amount of text from a web page, it links to that page.
• Images taken from the web show their source.

The format of answers also changed in 2026. Seer found that, at the same time citations dropped, Gemini started using headings in 99.5% of responses, up from about 3%. Tables appeared in 52% of responses, and the average answer got about 15% shorter. A short, structured answer leaves room for fewer sources.

Google-Extended, Googlebot and robots.txt

This is the biggest technical difference between Gemini and Google Search. Google AI Overviews and AI Mode depend only on Googlebot. The Gemini app also respects Google-Extended.

Google’s crawler documentation confirms that Google-Extended does not affect inclusion in Google Search and is not a ranking signal. So a site can block Google-Extended and still appear in AI Overviews, while losing grounded citations in the Gemini app. The user-triggered fetchers are listed separately.

Many sites block Google-Extended without knowing what it covers. Our study of SaaS robots.txt files found that 73% of top SaaS sites name no AI bot at all. Use our free robots.txt AI bot checker to see where you stand.

Which sites Gemini cites most in 2026

Ahrefs published the 50 most-cited sites in Gemini in September 2026, based on more than 3 million US queries. “Mention share” is each domain’s share of all citations going to the top 50.

Further down the list are more review and editorial sites: CNET (14th), Trustpilot (19th), Medium (20th), TechRadar (23rd), Reviewed (30th), the Better Business Bureau (35th) and Tom’s Guide (39th).

Reddit is far ahead. Google has had a data partnership with Reddit since February 2024 that gives it access to Reddit’s Data API to display and train on Reddit content.

Unlike Google AI Overviews, Gemini’s top 10 has no Facebook, Instagram or TikTok, and google.com is absent. It leans on reference sites, expert reviews and retailers instead. Profound’s data agrees: it describes Gemini as leaning editorial and reference, while AI Overviews and AI Mode lean social and user-generated. In Profound’s global data, G2 was the fifth most-cited domain in Gemini, which matters for software brands.

Gemini vs AI Overviews and AI Mode

All three come from Google and use Google Search, yet they name different brands. Profound tracked 15,155 brand setups daily in May 2026, about 1.48 million observations.

Gemini shared only about 27% of named companies with AI Overviews and 29% with AI Mode. The median brand had an 8-point visibility gap between its best and worst Google surface on the same day. Gemini gave the highest visibility in 50% of daily observations, more than either of the other two.

If you want the AI Overviews side in detail, see our guide on how Google AI Overviews and AI Mode choose sources. For comparison with OpenAI, read how ChatGPT chooses its sources.

Listicles, reviews and “best X” articles

“Best X” lists were a major source for Gemini, but 2026 brought a sharp change. In Seer’s data, citations of “best of” listicles fell 40% after the February shift. Medium’s share fell from 12.3% to 2.2%, and Forbes’s from 8.8% to 1.8%. Reddit and Wikipedia held steady at 44% and 33%.

The change hit commercial prompts hardest. For one ecommerce brand, the share of responses with citations fell from 96% to 3.7% in a week. Gemini moved from comparing products with sources to recommending products directly.

Two lessons follow. First, the lists that matter are those Gemini still reads: trusted review sites, expert publications and retailers that rank well. Second, recommendation answers that cite nothing come from what the model already knows. Being named on many credible lists over time feeds that knowledge, even if a single list is not cited today.

Freshness: does Gemini prefer new content?

Yes, moderately. An Ahrefs study of about 17 million citations across seven platforms (July 2025) found that Gemini’s cited pages averaged 1,118 days old. That made them 298 days newer than the organic results for the same queries. ChatGPT’s preference for new content was stronger, while Google AI Overviews showed no preference for new content at all.

Update pages when facts change: prices, features, rankings, statistics. Changing the date alone does not count as an update. Google’s John Mueller has warned against it.

Which brand signals matter for Gemini

There is less public research on Gemini than on ChatGPT or Google AI Overviews. The large Ahrefs study of 75,000 brands (December 2025), which found branded web mentions to be the strongest signal, did not cover Gemini. Treat the following as patterns from the source data, not measured correlations.

• Reddit presence. Reddit takes 28.5% of mention share in Ahrefs’ Gemini data and held steady through the 2026 citation drop.
• Wikipedia and reference sources. Wikipedia is third in Ahrefs’ list and held a steady 33% share in Seer’s data. Brands only get a page when independent sources show they are notable.
• Expert reviews. PCMag, CNET, TechRadar, Tom’s Guide and Edmunds all rank in the top 50. For software, G2 shows up in Profound’s data.
• Retail listings. Walmart and Target sit in the top 10, so product brands need complete and accurate retailer pages.
• Consistent facts. Since some answers come from training data, the same name, category and description everywhere makes it easier for the model to describe you correctly.

What a brand can do: a 9-step plan

1. Check your robots.txt for Google-Extended. If you block it, Gemini cannot ground answers in your pages. Decide on purpose whether that trade-off is worth it.
2. Make sure Google indexes your key pages. Grounding uses Google Search. Pages that do not rank cannot be cited.
3. Write down the prompts buyers use. Collect 30 to 50 real questions, including “best” and “top” prompts and how-to questions.
4. Answer the likely sub-questions. Cover pricing, alternatives, comparisons and use cases on clear pages. Put the answer in the first sentence under each heading.
5. Get onto review sites and lists Gemini cites. Run your prompts, note the cited domains and target those. Our brand mention service handles the outreach.
6. Take part honestly on Reddit. Answer questions in relevant subreddits, disclose who you are and follow each community’s self-promotion rules.
7. Publish useful YouTube videos. Clear titles, descriptions and transcripts help. YouTube is second in Ahrefs’ Gemini list, even though Seer saw its share fall in 2026.
8. Earn coverage in trusted publications. Original data gives journalists a reason to write about you. See our digital PR and data-driven content services.
9. Track monthly, cited and uncited. Log whether each answer had sources, which domains were cited and which brands were named. Compare with your top three competitors.

Limits: what no one can guarantee

• Not every answer searches. In Seer’s sample, about a quarter of Gemini responses had no citations after March 2026. Those answers come from training data.
• Behavior changes without notice. The 2026 drop in citations, the new heading format and the fall in YouTube citations from 18% to 3% of responses were not announced in advance.
• The app and the API differ. The grounding steps above come from Google’s developer documentation. The Gemini app may behave differently.
• Few answer slots. With about 6.6 citations per run, Gemini has fewer places for your pages than AI Overviews or AI Mode.
• Studies are correlational and short-lived. They describe what Gemini cited in a given period, not what causes a citation.

A realistic goal is to raise the share of relevant prompts where Gemini names your brand, then track it over months against competitors.

FAQs

Does Gemini use Google Search to find sources?

Yes, when it grounds an answer. Google’s developer documentation says the model decides whether a search would help, writes one or more queries, reads the results and adds citations. When it does not search, Gemini answers from training data and shows no sources. Not every Gemini response includes sources.

If I block Google-Extended, will Gemini stop citing my site?

It should stop using your content for grounding in the Gemini app and Vertex AI, and for training future Gemini models. Google says Google-Extended does not affect Google Search, so AI Overviews and AI Mode can still cite you. User-triggered fetchers, such as Google-GeminiNotebook, generally ignore robots.txt.

Why does Gemini sometimes give no sources?

Gemini only adds sources when it searches or quotes a web page. Seer Interactive found the share of cited responses fell from 99% to 76% in early 2026, mostly on “best” and “top” recommendation prompts. Those answers draw on training data, so long-term brand mentions across the web still matter.

Is Gemini the same as Google AI Overviews?

No. Both use Google Search, but they cite different sites and name different brands. Profound found Gemini shares only about 27% of named companies with AI Overviews. Gemini cites fewer sources per answer and leans on editorial and reference sites, while AI Overviews lean on YouTube and social platforms.

Which sites does Gemini cite most often?

In Ahrefs’ September 2026 data from more than 3 million US queries, Reddit led with 28.5% mention share, followed by YouTube at 14.6% and Wikipedia at 8.8%. Retailers such as Walmart and Target, plus review sites like PCMag, Edmunds and Kelley Blue Book, also ranked in the top 10.

Can I pay to be recommended by Gemini?

We found no verified way to buy a place in Gemini’s answers. Google’s documentation describes grounding as selecting sources from Google Search results. Brands get named because they appear on pages Gemini finds and trusts, such as reviews, discussions, reference sites and press coverage, not through payment.

Get your brand into the sources Gemini reads

Gemini names the brands it keeps finding in discussions, reviews, reference sites and press coverage. Create & Grow places brands in relevant “best X” articles, reviews and press pieces, and runs link building and digital PR campaigns that earn mentions on sites AI assistants cite. Brand mention plans start at €1,500 a month.

We cannot guarantee that Gemini will mention you, and nobody honestly can. We can show you which sources Gemini cites for your category and get you into the ones that matter. Contact us to start with a short audit of your prompts.

More guides in this series on how AI search picks sources: ChatGPT, Google AI Overviews, Perplexity, Claude, Copilot.

That makes Google’s AI answers closer to classic SEO than ChatGPT’s. Organic rankings still matter, but they are not the whole story. In 2026, more than half of the pages cited in AI Overviews ranked outside the top 10 for the main query. AI Mode overlaps even less with the regular results. Brands get named when they keep showing up across the pages Google reads: YouTube videos, Reddit threads, “best X” lists, reviews and their own Google Business Profile.

Google does not sell citations in AI Overviews or AI Mode, and nobody can promise you one. You can raise the odds. The 2025 and 2026 data below shows where those odds come from.

• The short answer: what decides a Google AI citation
• AI Overviews vs AI Mode: same index, different answers
• How Google finds and cites pages, step by step
• How citations and links appear in AI answers
• Googlebot, Google-Extended and robots.txt
• Which sites Google’s AI cites most in 2026
• Do you need to rank in the top 10?
• Listicles, reviews and “best X” articles
• Freshness: does Google’s AI prefer new content?
• Which brand signals correlate with Google AI mentions
• What a brand can do: a 9-step plan
• Limits: what no one can guarantee
• FAQs
• Get your brand into the sources Google’s AI reads

The short answer: what decides a Google AI citation

Here is the whole process in one table. Each row is covered in detail below.

AI Overviews vs AI Mode: same index, different answers

AI Overviews are the AI summaries at the top of a normal results page. AI Mode is the separate, chat-style search tab built for longer questions and follow-ups. Google’s Search Central documentation says the two may use different models and techniques, so the answers and the links can differ.

Google is also merging them. In the Q2 2026 earnings remarks, Sundar Pichai said AI Overviews and AI Mode had recently been brought together into “one seamless Search experience”. He also said AI Mode had passed 1 billion monthly users.

Third-party data shows how far apart the two still are:

The Moz figures come from Tom Capper’s February 2026 study of 10,000 keywords, each checked on desktop and mobile in one US and one UK location. The Profound figures come from its July 2026 analysis of 15,155 tracked brands. Profound found that AI Overviews and AI Mode share only about 40% of the companies they name for the same prompts.

In practice, treat them as two channels. A brand can show up well in one and barely at all in the other.

How Google finds and cites pages, step by step

1. Google decides whether to show an AI answer

AI Overviews do not appear on every search. Google says they show only when they add something beyond the normal results. An Ahrefs study of 146 million results pages (January 2026) found AI Overviews on 21% of keywords, 57.9% of question queries and 46.4% of queries with seven or more words. In that data, 99.9% of keywords that triggered an AI Overview were informational.

A Pew Research Center study of 900 US adults and their real searches in March 2025 found a similar pattern. 18% of searches produced an AI summary. That rose to 60% for searches starting with a question word and 53% for searches of 10 words or more.

2. Google fans the question out

Google described query fan-out at I/O 2025: AI Mode breaks a question into subtopics and runs many searches at once. Deep Search, a mode inside AI Mode, can run hundreds of searches and return a fully cited report. Google’s documentation says both AI Overviews and AI Mode may use fan-out across subtopics and data sources.

This is why a page can be cited for a query it does not rank for. It may rank for one of the hidden sub-questions instead.

3. Google reads from its index and its own data

The candidate pages come from Google’s search index. Google also uses its own structured data. Its documentation asks businesses to keep Merchant Center and Business Profile information up to date. Profound found that google.com’s share of AI Mode citations rose 8.4 times between April 15 and June 30, 2026, making it the second most-cited domain. The growth came from Business Profiles and product knowledge panels, mainly in hospitality, home services, restaurants, real estate and healthcare.

4. Google links the pages that support the answer

The model writes the answer and attaches supporting links. Google says AI features show “a wider and more diverse set of helpful links” than a classic results page. Pew found that 88% of AI summaries cited three or more sources.

How citations and links appear in AI answers

Links show up in a few places: inline next to a statement, in a source panel beside the answer, and in lists of related articles. In October 2026, Google announced several changes:

• More inline links, placed higher up in responses.
• A test of a dedicated section at the top for official websites, with context about each.
• News carousels that highlight a user’s Preferred Sources.
• Thumbnails on related-article links.
• Clickable creator and publisher names in AI Mode that open their Search profiles.

Clicks are the weak point. In the Pew study, users clicked a regular result in 8% of visits with an AI summary, against 15% without one. They clicked a link inside the summary in just 1% of visits. Google’s side of the story is different. Pichai said Google sends “billions of clicks to websites every week” through AI features, and Google’s documentation calls those clicks “higher quality”.

You can measure this traffic in Search Console. Google says AI feature clicks and impressions are counted in the Performance report under the “Web” search type. They are not broken out separately.

Googlebot, Google-Extended and robots.txt

This is where Google differs most from OpenAI. There is no separate “AI search bot” for AI Overviews or AI Mode. Both run on what Googlebot crawls for Search.

The Google crawler documentation confirms that Google-Extended has no separate user agent. It only tells Google whether content may be used to train Gemini models and to ground answers in the Gemini app and Vertex AI. Blocking it does not keep you out of AI Overviews.

Google also runs fetchers that act on a user’s request, such as Google-Agent. Google says these generally ignore robots.txt because a person asked for the fetch.

Check your setup before anything else. Our study of SaaS robots.txt files found that 73% of top SaaS sites name no AI bot at all. Our free robots.txt AI bot checker shows what your file allows.

Which sites Google’s AI cites most in 2026

Ahrefs published separate top-50 lists for AI Overviews and AI Mode in September 2026, based on more than 3 million US queries across all topics. “Mention share” is each domain’s share of all citations going to the top 50.

Three things stand out:

• Video and social lead. YouTube, Reddit, Facebook, Instagram and TikTok take most of the top spots. Google leans far more on user-generated content than ChatGPT does. Google also has a data deal with Reddit from 2024 that lets it display and train on Reddit content.
• Google cites itself. google.com is fourth in AI Overviews and third in AI Mode, mostly through Business Profiles and product panels.
• Publishers and review sites sit lower. Forbes is 12th in both lists. Trustpilot is 35th in AI Overviews and 44th in AI Mode. Consumer Reports is 30th in AI Overviews.

Pew’s 2025 data points the same way. Wikipedia, YouTube and Reddit together made up 15% of the sources in AI summaries. Government (.gov) sites made up 6% of AI summary sources, against 2% of standard results.

These are averages across every topic. In a B2B software niche, the cited pages are more likely to be vendor sites, review platforms and “best X” articles. Check your own prompts before you copy the averages.

Do you need to rank in the top 10?

It helps, but it is not required. The studies disagree on how much, mostly because they measure different things.

Conductor’s larger and newer dataset also shows how the odds change by position. A page ranking first was cited 24.9% of the time. That fell to 21.1% at position two, 17.9% at position three and 10.2% at position ten.

Si Quan Ong of Ahrefs summed it up well: ranking first makes a citation more likely, “but that chance is a coin flip at best.” Fan-out explains the rest. Pages that rank for the sub-questions get pulled in even when they are invisible for the main query.

Listicles, reviews and “best X” articles

“Best X” lists carry a lot of weight in Google’s AI answers. An Ahrefs study by Glen Allsopp (December 2025) looked at 750 commercial prompts about software, products and agencies. Blog lists of the “best X” made up 43.8% of the page types ChatGPT cited, and they were slightly more prominent in AI Overviews. Ahrefs later charted the share by platform:

The same study found that 79.1% of the 1,100 cited lists had been updated in 2025, and 35% sat on low-authority domains. In Google’s regular results, 169 of 250 “best X software” results pages included a list where the author ranked their own product first.

So being on the right lists matters more than domain strength alone. Lists that are kept current and explain why each product is included are the ones worth targeting. A list where a vendor ranks itself first can still get cited, but readers trust it less.

Freshness: does Google’s AI prefer new content?

Less than other AI assistants. An Ahrefs study of about 17 million citations across seven platforms (July 2025) found that AI assistants cite content that is 25.7% newer than organic results on average. AI Overviews were the exception. They cited the oldest content of all the platforms, about 1,432 days old at publication, slightly older than the organic results.

This fits how AI Overviews work. They lean on the same pages that already rank, and many of those are established. Update a page when the facts change, not to reset a date. Google’s John Mueller has warned against changing dates without changing the content.

Which brand signals correlate with Google AI mentions

Ahrefs studied 75,000 brands (December 2025) and measured which signals best tracked how often each brand was mentioned in AI answers. The numbers are Spearman correlations: 1.0 is a perfect match, 0 is no relationship.

Mentions on YouTube showed the strongest correlation of any signal in the study, at about 0.737. Ahrefs called AI Mode a “consensus engine”: it names brands that many sources agree on, which makes it the hardest of the three for new brands.

The brands each assistant named also overlapped heavily. The correlation between AI Overviews and AI Mode was 0.821. Work that gets you named in one Google surface tends to help in the other.

These are correlations, not proof of cause. Brands with many mentions also tend to be bigger, older and better known. Still, the pattern is consistent: being talked about on other sites lines up with AI visibility far better than domain authority does.

What a brand can do: a 9-step plan

1. Confirm Google can index your key pages. Check robots.txt, noindex tags and snippet controls on product, pricing and comparison pages. Do not block Googlebot. Blocking Google-Extended is a separate choice that does not affect AI Overviews.
2. Write down the prompts buyers use. Collect 30 to 50 real questions from sales calls, support tickets and People Also Ask. Include long, specific questions, since those trigger AI answers most often.
3. Map the fan-out. For each prompt, list the sub-questions Google is likely to search: pricing, alternatives, integrations, use cases, comparisons. Each one is a page or a section you need.
4. Answer each sub-question in a self-contained passage. Put the answer in the first sentence under a clear heading. Keep the key facts in text, not only in images or tabs.
5. Complete your Google Business Profile and Merchant Center feed. AI Mode now cites google.com heavily through these. Keep hours, categories, products and prices accurate, and reply to reviews.
6. Get onto the “best X” lists Google already cites. Run your prompts, note the cited lists and contact their authors with a clear reason to include you. Our brand mention service does this for you.
7. Show up on YouTube and Reddit. Publish useful videos with clear titles and transcripts. Take part honestly in relevant subreddits and follow each community’s rules on self-promotion.
8. Earn coverage with original data. Surveys and studies give journalists and list authors a reason to mention you. See our digital PR and data-driven content services.
9. Track it every month. Run the same prompts in AI Overviews and AI Mode, log the cited URLs and named brands, and compare with your top three competitors. Use Search Console for clicks.

Much of this overlaps with what works for ChatGPT. Our guide on how ChatGPT chooses its sources covers the differences in detail.

Limits: what no one can guarantee

• No paid route into the answer. Google says no special optimization is required for AI features, and there is no way to pay for a citation inside an AI Overview or AI Mode answer.
• Answers vary. Profound found a median gap of 8 visibility points between a brand’s best and worst Google AI surface on the same day. More than one in three brands had gaps above 10 points.
• Google changes the interface often. Inline links, official-site sections and Preferred Sources all changed in 2026. The data above may look different in six months.
• Fewer clicks. Even a citation may bring little traffic. In the Pew study, users clicked a link inside the AI summary in 1% of visits.
• Most studies are correlational. They show what cited pages and brands have in common, not what causes a citation.

A realistic goal is to raise the share of relevant prompts where your brand appears, then track it over months against competitors.

FAQs

Do I need to rank in the top 10 to appear in AI Overviews?

No, but it helps. Conductor’s 2026 data shows 43% of AI Overview citations rank in the organic top 10, and a first-place page is cited about a quarter of the time. The rest come from pages that rank for related sub-questions Google searches behind the scenes through query fan-out.

Does blocking Google-Extended remove me from AI Overviews?

No. Google says Google-Extended only controls whether your content trains Gemini models and grounds answers in the Gemini app and Vertex AI. It does not affect Search or AI Overviews. To keep content out of AI Overviews, use nosnippet, data-nosnippet, max-snippet or noindex, which also limit regular Search.

Is optimizing for AI Overviews different from regular SEO?

Google says no special optimization is needed beyond normal SEO best practices. The difference is scope. Fan-out means you need pages for the sub-questions, and brand mentions on YouTube, Reddit and “best X” lists influence which brands Google names. Classic rankings are necessary but rarely enough.

Why does AI Mode cite different sites than AI Overviews?

Google says the two may use different models and techniques. AI Mode runs more searches and cites more sources per answer, about 15 versus 11 in Profound’s 2026 data. Moz found AI Mode’s cited URLs overlap with the organic top 10 only 12% of the time, against 88% for AI Overviews.

Can I see AI Overview traffic in Google Search Console?

Partly. Google counts clicks and impressions from AI Overviews and AI Mode in the Search Console Performance report under the “Web” search type. They are mixed in with regular results, not shown as a separate filter. To see which prompts cite you, you need manual checks or an AI visibility tool.

Does structured data help my page get cited?

Google lists structured data as a best practice, as long as it matches the visible text on the page. It is not a special AI signal and does not guarantee a citation. It helps Google understand products, prices and reviews, and keeping Merchant Center and Business Profile data accurate matters too.

Get your brand into the sources Google’s AI reads

Google’s AI names the brands it keeps finding in videos, discussions, lists, reviews and press coverage. Create & Grow places brands in relevant “best X” articles, reviews and press pieces, and runs link building and digital PR campaigns that earn mentions on sites AI assistants cite. Brand mention plans start at €1,500 a month.

We cannot guarantee that AI Overviews or AI Mode will mention you, and nobody honestly can. We can show you which sources Google cites for your category and get you into the ones that matter. Contact us to start with a short audit of your prompts.

More guides in this series on how AI search picks sources: ChatGPT, Gemini, Perplexity, Claude, Copilot.

For brands, the second path is the one you can influence. A page has to be crawlable by OpenAI’s search bot, show up in the search results ChatGPT pulls, and match the sub-questions ChatGPT asks behind the scenes. Brands then get named when they appear often and consistently across the pages ChatGPT reads: “best X” lists, review sites, Reddit threads, Wikipedia, the press and their own product pages.

OpenAI says plainly that placement in ChatGPT search is not guaranteed. Nobody can promise a mention. What you can do is raise the odds, and the 2026 data below shows where those odds come from.

• The short answer: what decides a ChatGPT citation
• Training data vs live search: two ways ChatGPT knows your brand
• How ChatGPT search finds and cites pages, step by step
• How citations and brand links appear in answers
• OpenAI’s crawlers and robots.txt
• Which sites ChatGPT cites most in 2026
• Listicles, reviews and “best X” articles
• Freshness: does ChatGPT prefer new content?
• Which brand signals correlate with ChatGPT mentions
• What a brand can do: a 9-step plan
• Limits: what no one can guarantee
• FAQs
• Get your brand into the sources ChatGPT reads

The short answer: what decides a ChatGPT citation

Here is the whole process in one table. Each row is covered in detail below.

Training data vs live search: two ways ChatGPT knows your brand

ChatGPT can talk about your brand without searching at all. In that case it relies on what the model learned during training. There is no citation, no link and no way to see which pages shaped the answer.

OpenAI separates the two paths in its crawler setup. According to OpenAI’s crawler documentation, GPTBot collects content that may be used to train its foundation models. OAI-SearchBot is used to show websites in ChatGPT’s search features. The two settings are independent.

Live search is where citations come from. OpenAI’s help page on ChatGPT search says ChatGPT may search automatically when a question would benefit from current information, and users can also start a search manually.

Most marketing questions (“best CRM for startups”, “alternatives to X”) are the kind that trigger a search. That is why the rest of this guide focuses on the search path.

How ChatGPT search finds and cites pages, step by step

1. ChatGPT rewrites the prompt into search queries

OpenAI explains that ChatGPT search “sometimes partners with other search providers” and typically rewrites the user’s question into one or more targeted queries. After it reviews the first results, it may send more specific follow-up queries. It can also use approximate location and, if memory is on, saved memories when rewriting the query (OpenAI).

The scale of this is large. In an AirOps study published in March 2026 (15,000 prompts, 548,534 retrieved pages), ChatGPT generated two or more fan-out queries on 89.6% of searches. 95% of those fan-out queries had zero monthly search volume in traditional keyword tools.

For comparison prompts, the fan-out often splits the question apart. AirOps gives the example of “HubSpot vs Salesforce” turning into separate searches about pricing, features and reviews.

2. Search partners return candidate pages

OpenAI does not publish a full list of its search providers. Its help page links to the privacy statements of Microsoft and Shopify as providers that may process queries. In practice, this is why many SEOs check that a site is indexed well in Bing, not only Google.

Google rankings still matter, even if Google is not named as a provider. In the AirOps data, pages ranking #1 in Google were cited by ChatGPT 43.2% of the time. That was 3.5 times the citation rate of pages ranking outside Google’s top 20. This is a correlation: pages that rank well on Google usually rank well elsewhere too.

3. ChatGPT screens results before it reads them

An Ahrefs study of 1.4 million prompts (published April 2026) describes a gatekeeping layer. Each search result arrives with a title, URL, sometimes a snippet and an ID. ChatGPT uses that data to decide which pages to open.

Ahrefs found that cited pages had titles that were more semantically similar to the prompt and to the fan-out queries than non-cited pages. Search results with natural-language URL slugs had an 89.78% citation rate, compared with 81.11% for those without.

4. ChatGPT cites only part of what it retrieves

Both large studies agree that retrieval is not the same as citation. They disagree on the exact share, because they count retrieved pages differently.

Ahrefs also found that 88% of cited URLs came through ChatGPT’s general “search” channel. Reddit content pulled in through a separate channel made up 67.8% of the URLs that were retrieved but not cited. In other words, ChatGPT reads a lot of Reddit for context but cites it much less often than it reads it.

What gets a retrieved page cited? AirOps saw a citation rate of 20.1% for pages whose title overlapped with the query by 50% or more, against 9.3% for pages with less than 10% overlap. Easier-to-read pages were also more common among cited sources.

5. The same prompt can give different sources

ChatGPT’s sources shift between modes and over time. In a Semrush analysis with Kevin Indig (100 prompts across 20 buyer journeys, reported by Search Engine Land), only 25.6% of cited domains overlapped between minimal and high reasoning for the same prompts. High reasoning ran 1,130 web searches across the test set, against 245 for minimal reasoning.

The mix of sources changed too. With high reasoning, Reddit’s citation share fell from 15% to 7%, and official documentation and support pages rose from 12.4% to 17.5%. A brand that shows up in quick answers can disappear in deeper research answers, and the reverse.

How citations and brand links appear in answers

According to OpenAI, responses that use web search “may include citations”. Users can click a citation to open the source, hover over it on desktop to preview it, or open Sources to see cited sources and other relevant links (OpenAI). OpenAI also warns that citations can be incomplete, outdated or wrong.

The May 2026 branded link change

There is now a difference between a citation (a link to the page that supports a claim) and a brand link (a link on the brand name itself). Profound reports that on May 7, 2026 ChatGPT began turning brand names in answers into links to the brand’s homepage far more often (Profound).

• The share of ChatGPT responses containing a URL went from about 4.5% to 20–24% on the same day.
• Homepages went from about 4% to 24% of OpenAI referrals across the sites Profound monitors.
• B2B software and SaaS brands saw the largest gain, with daily OpenAI referrals up more than 200% on their earlier baseline. E-commerce stayed roughly flat.

OpenAI did not announce this change, so treat these as one vendor’s measurements. The practical point holds anyway: being named in an answer can now send traffic, not just awareness. Your homepage has to work for visitors who arrive knowing only what ChatGPT said about you.

OpenAI’s crawlers and robots.txt

This is the one part of the process you fully control. OpenAI lists four user agents in its crawler documentation.

Key facts from OpenAI’s documentation:

• Sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links.
• You can allow OAI-SearchBot and block GPTBot. That keeps you eligible for search answers while signalling you do not want your content used for training.
• Changes to robots.txt can take about 24 hours to take effect for search.
• OpenAI also asks site owners to make sure their host or CDN allows traffic from its published OAI-SearchBot IP addresses. A firewall that blocks the bot cancels out a correct robots.txt.

Many sites have never made this decision. Our study of AI bots in SaaS robots.txt files found that 73% of top SaaS sites name no AI bot at all. You can check your own file with the free robots.txt AI bot checker on our tools page.

What about llms.txt? OpenAI’s crawler documentation does not mention it. Adoption is growing anyway: 74% of 47 leading B2B SaaS sites serve an llms.txt file. Treat it as optional housekeeping, not a ranking lever.

Which sites ChatGPT cites most in 2026

The most-cited domains tell you where ChatGPT looks when it builds an answer. If your brand is absent from those places, it is less likely to be named.

The freshest public snapshot is Ahrefs’ list of the 50 most-cited websites in ChatGPT, based on broad US queries across all topics in September 2026.

Three things stand out:

• Reddit and Wikipedia lead, but the lead moves. A Semrush study of 230,000 prompts (July to October 2025) found ChatGPT cited Reddit in close to 60% of responses in early August 2025, then around 10% by mid-September. Wikipedia fell from roughly 55% of responses to under 20% in the same period.
• Review and comparison sites are a big share. Consumer Reports, TechRadar, Trustpilot, Tom’s Guide and similar sites sit in the top 50.
• The list is consumer-heavy. It covers all topics, so B2B sources look smaller than they are in a SaaS category. In Profound’s analysis of 680 million citations (August 2024 to June 2025), G2 took 1.1% of all ChatGPT citations, the same as Forbes.

The lesson for a SaaS brand: find the sources ChatGPT cites for your prompts. A general top-50 list is a starting point, not a target list.

Listicles, reviews and “best X” articles

“Best X” articles are the most direct route to a brand mention, because they already have the shape of the answer ChatGPT is building: a shortlist of named options with reasons. They are also the format that changed most in 2026.

What the 2026 data says

• Listicles are a large share of citations. Seer Interactive analysed 2 million ChatGPT citations from November 2025 to February 2026. URLs containing “best” or “top” made up 16.6% of them.
• Self-serving lists are losing ground. In the same data, listicle citations fell 30% from December 2025 to January 2026, faster than the 22.7% drop in total citations. 13 of 16 industries saw declines. Wikipedia, Reddit and review sites such as G2 and Capterra gained share.
• Brand product pages are rising. According to David Konitzny’s analysis of the GPT-5.6 rollout in August 2026, summarized by Peec, listicles fell from 15.47% to 8.19% of retrieved pages on transactional prompts. Product pages led at 16.39%, usually on the brand’s own domain, driven by fan-out searches that use the site: operator.

Which lists still get cited

Seer looked at 78 listicles that gained citations while the rest declined. They shared four traits:

1. Fresh. Recently updated pages grew. Older lists that had been cited for years dropped.
2. Comprehensive. Growing pages listed 10 to 20 options. Declining ones often listed 5 to 8 with thin comparisons.
3. Backed by evidence. They used outside data such as G2 ratings or a disclosed methodology.
4. Structured. Every one used numbered lists, often with features, pros and cons, pricing and “best for” sections.

So the goal is not “get on any list”. It is to be included, with accurate detail, in independent, well-maintained comparisons that rank for the queries ChatGPT runs. That is what our brand mentions service focuses on.

Reddit, Wikipedia and review sites

These three source types play different roles:

• Reddit is retrieved constantly and cited less often than it is read, according to Ahrefs. It shapes ChatGPT’s sense of what real users recommend. Seer suggests checking Reddit threads in your category to see which brands users mention and whether your own messaging matches.
• Wikipedia is a top-two domain in every dataset above. Most brands do not qualify for an article, and Wikipedia’s conflict-of-interest rules discourage brands from editing their own pages. Focus on the press coverage that would make you notable instead.
• Review sites (G2, Capterra, Trustpilot, Consumer Reports) feed both citations and the “external validation” that the best listicles quote. Up-to-date profiles with recent reviews are basic hygiene.

Freshness: does ChatGPT prefer new content?

Yes, but freshness is not enough on its own. Ahrefs reports that in an earlier study of 17 million citations, ChatGPT cited URLs that were 458 days newer than Google’s organic results. In its 1.4-million-prompt study, the median cited page was still about 500 days old (Ahrefs).

Within one set of retrieved results, the newest pages were often the ones left out. Relevance did the heavy lifting. Freshness acted as a tie-breaker, most clearly for news queries, where cited pages were younger than non-cited ones.

In practice:

• Update comparison and pricing pages when facts change, not just the date.
• Expect freshness to matter most for “best X in 2026”, news and pricing prompts.
• Do not expect a new page to beat a relevant, established one just because it is new.

Which brand signals correlate with ChatGPT mentions

Citations are about pages. Mentions are about brands. Ahrefs studied 75,000 brands (December 2025) to see which signals correlate with being mentioned in ChatGPT, Google AI Mode and AI Overviews.

Ahrefs stresses that correlation is not causation. Still, the pattern is consistent: brands that are talked about widely, in articles, videos and forums, get named more. Link counts and content volume on their own show weak relationships.

Ahrefs also found that ChatGPT correlates less with classic authority metrics than Google’s AI features do. That may make ChatGPT a more open door for smaller brands. The AirOps data points the same way: about 74% of citations went to sites with a Domain Authority under 80.

This is why digital PR and earned mentions sit at the centre of AI visibility work. Links still help pages rank in the search results ChatGPT pulls from, so link building supports the retrieval step. But the brand mention itself is what gets you named.

What a brand can do: a 9-step plan

1. Allow OAI-SearchBot. Check robots.txt and your CDN or firewall rules. Decide separately whether to allow GPTBot.
2. Get indexed beyond Google. Confirm your key pages are indexed in Bing, since Microsoft is one of the search providers OpenAI names.
3. Build a prompt list. Write 20 to 50 prompts a buyer would ask: “best [category] for [use case]”, “[competitor] alternatives”, “[your brand] vs [competitor]”, “[your brand] pricing”.
4. Record who gets cited. Run the prompts (logged out or with memory off, in more than one mode) and note the cited URLs and named brands. Repeat monthly, because results move.
5. Fix your own pages. Make product, pricing, feature and comparison pages clear, current and readable without JavaScript. Use titles and URLs that match how people ask.
6. Cover the fan-out questions. Answer pricing, integrations, security and “who it is for” questions directly on your site. Fan-out queries often target exactly these.
7. Get into the lists ChatGPT already cites. Pitch the independent “best X” articles and reviews from step 4 with accurate, useful information. Prefer lists that are maintained and show their method.
8. Earn press and expert mentions. Original data gives journalists a reason to name you. Data-driven content and digital PR campaigns create mentions on sites ChatGPT trusts.
9. Keep third-party profiles consistent. Update review profiles, LinkedIn, directories and YouTube descriptions so your name, category and positioning match everywhere.

Limits: what no one can guarantee

Be careful with anyone who promises ChatGPT placements. The facts do not support it.

• OpenAI does not guarantee placement. Its help page says ChatGPT ranks results “using multiple factors” and that “placement is not guaranteed”.
• Answers vary. Location, memory, model and reasoning mode change the queries ChatGPT runs and the sources it cites. Only a quarter of cited domains overlapped between reasoning modes in the Semrush study.
• Source preferences shift without notice. Reddit and Wikipedia citations dropped sharply in September 2025. Listicles dropped in early 2026. Brand links jumped in May 2026. None of these changes were announced in advance.
• Training data is slow. Work that affects how the model itself describes you only shows up after future model updates.
• Most studies are correlational. They show what cited pages have in common, not what causes a citation.

A sensible goal is to raise the share of relevant prompts where your brand appears, track it over months and compare it with competitors. A single screenshot proves little in either direction.

FAQs

Does ChatGPT use Google or Bing for search?

OpenAI says ChatGPT search partners with third-party search providers and links to Microsoft’s and Shopify’s privacy policies. It does not publish a full provider list. Google rankings still correlate with ChatGPT citations, so it makes sense to be indexed and ranking well in both Google and Bing.

If I block GPTBot, will ChatGPT stop mentioning my brand?

No. GPTBot controls whether content may be used for model training. OAI-SearchBot controls whether your site can appear in ChatGPT search answers. You can block GPTBot and still allow OAI-SearchBot. Blocking OAI-SearchBot keeps your pages out of search answers, except as possible navigational links.

Can I pay to be recommended by ChatGPT?

ChatGPT now carries ads, and OpenAI uses a separate bot, OAI-AdsBot, to check ad landing pages. An ad is not the same as being named in the answer. We found no verified way to buy a place in ChatGPT’s organic answers. Mentions come from what ChatGPT finds across the web.

How long does it take for ChatGPT to cite a new page?

There is no fixed timeline. OpenAI says robots.txt changes can take about 24 hours to reach its search systems. After that, a page still needs to rank for the queries ChatGPT runs and answer them better than competing pages. Being crawlable makes a page eligible, not cited.

Why does ChatGPT give different brands for the same question?

Each run can rewrite the question into different searches and pull different results. Location, memory and the reasoning mode also change the sources. In a Semrush test with Kevin Indig, only 25.6% of cited domains overlapped between quick and deeper reasoning answers to the same prompts. Track trends over many runs.

Does llms.txt help my brand appear in ChatGPT?

OpenAI’s crawler documentation does not mention llms.txt, and we found no evidence that ChatGPT uses it to choose sources. It does no harm and many SaaS sites already serve one. The proven basics are allowing OAI-SearchBot, ranking for relevant queries and earning mentions on sites ChatGPT cites.

How can I track whether ChatGPT mentions my brand?

Run a fixed list of buyer prompts every month and log the cited URLs and named brands. AI visibility tools from Ahrefs, Semrush, Profound and others automate this at scale. Also watch referral traffic from chatgpt.com in your analytics, especially since brand names started linking to homepages in May 2026.

Get your brand into the sources ChatGPT reads

ChatGPT names the brands it keeps finding in trusted lists, reviews and press coverage. Create & Grow places brands in relevant “best X” articles, reviews and press pieces, and runs digital PR campaigns that earn mentions on sites AI assistants cite. Brand mention plans start at €1,500 a month.

We cannot guarantee that ChatGPT will mention you, and nobody honestly can. We can show you which sources it cites for your category and get you into the ones that matter. Contact us to start with a short audit of your prompts.

More guides in this series on how AI search picks sources: Google AI Overviews, Gemini, Perplexity, Claude, Copilot.

Your link request is sitting in an inbox nobody opens

I have sent a lot of cold emails to editors. Most of them went nowhere, and not because the pitch was bad. The editor got 200 messages that week, and mine was one more subject line sliding past the fold.

Link building and partnership outreach fail in the inbox for a simple reason. Everyone uses email, so every inbox looks the same. A stranger with a link request lands in the same pile as the newsletters, the vendor invites, and the ‘quick question’ follow-ups.

So the question I keep asking is not ‘how do I write a better email?’ It is ‘where does this person already pay attention?’ For a lot of editors, webmasters, and partner marketers, the answer is LinkedIn. They post there, they comment there, and they read direct messages there, often from people they recognize.

This article covers how to run link and partnership outreach on LinkedIn without turning your account into a spam machine. I will show you when the channel makes sense, how to structure three touches that do not beg for a link in the first message, how to scale the work safely, and how to tell when a campaign is failing and why.

When LinkedIn beats the inbox, and when email still wins

An editor who publishes under their own name, a webmaster who posts about their projects, and a marketer who runs everything in a startup are all easier to reach out on Linkedin than in a generic contact form on their website.

When choosing who to reach out to on Linkedin, I start from the company page, and see all the employees. If there is someone in charge of SEO, then that is the person I contact first, if there is not, then I try to find editors or marketers. They are either handling SEO on the side, or have freelancers who are in charge of backlinks.

But before you start doing Linkedin outreach, your profile needs to look legit. You have to have your picture, clear description of what you do, and more than 500 connections so you don’t look like a fake profile. Also, I would suggest you to skip writing the note when sending a connection request, as that one might backfire if it is not written well.

The main advantage of Linkedin is that it feels more personal, as it is a social media platform after all. And I think that notification on Linkedin is much harder to ignore, when it comes with a real face and real question.

Research before you request

My process starts with competitors’ backlinks, not with LinkedIn. I pull the backlink profiles of the competitors I care about in Ahrefs and export them. Every site that links to a competitor already covers the topic, which makes it a warm lead for a link.

Then I clean the export. I remove duplicates and the domains that clearly do not fit. Next, I compare the list against the sites that already link to us, because there is no point asking for a link we already have. What remains is the target list.

The enrichment work happens in Clay, handled by our GTM engineers. For each site, they find the name, surname, LinkedIn profile, position (SEO, editor, marketer) and email of the person behind it. If I cannot match a site to a real person, I drop it instead of guessing.

Personalization comes next, and I use it only when I have a concrete reason. My last campaign went to sites that had mentioned a tool that is no longer on the market. The message said the tool was gone and asked whether they would mention mine instead, and it linked to the article where they had written about the old one. The reason was real, the link was specific, and the ask was easy to understand.

I run these campaigns in HeyReach because it handles the parts I would otherwise do by hand. It stops the same person from being messaged twice. It lets me add people and websites to a blocklist so I never contact them. And when people do not reply, I can pull them through the HeyReach MCP into a new list for a simple follow-up a month or two later.

Four touches that do not pitch on touch one

My sequence has three touches. In my campaigns, the ask can come earlier when it rests on a concrete reason, like the example above. Otherwise, I wait until the person replies before I make the ask.

Here is how the branches look like for my campaigns.

The first touch is the connection request. I keep it short and specific, or I skip the note altogether. Nobody accepts a request because it contains a paragraph of hope. In HeyReach’s benchmark report, we found that the typical campaign gets about one accepted connection for every five requests sent, so acceptance is a fair thing to expect but not a guarantee you can bank on.

The second touch is an observation. Once the request is accepted, I comment on one specific piece of the person’s work and ask one question about it. This is the step most campaigns get wrong, and it is where I put the most care.

The third touch is the ask. I send it only after the person replies, and I offer something concrete: a data point they can use, a case study that fits their audience, or a specific angle for a piece they are working on. If they prefer email from there, I move the conversation over.

The fourth touch is a follow up. I never say bumping this on top of your inbox, or anything like that. I tell them I understand that they are busy, and just tell them how I feel. Sometimes it is a meme, sometimes it is a pop culture reference.

Scaling the touches without breaking your account

Scale feels like the goal, but scale is where careful outreach turns into noise. I learned that the hard way. When I sent too many messages by hand, it was so daunting, and I also couldn’t always keep track of how many requests I was sending per day. Automation tools help you pace the sending. Some tools on the market set caps for you, so check what each one enforces before you trust it with your account. No tool guarantees your account is safe, so stay inside LinkedIn’s limits (100 messages per week for free accounts ).

In HeyReach’s benchmark report, acceptance drifted down as campaigns grew. Median acceptance went from 21.43% in campaigns with 50 to 100 requests to 19.35% in campaigns with more than 1,000. Reply rates stayed roughly flat over the same range. So bigger campaigns do not kill conversations. They tend to reveal list quality problems upstream. If your large campaign underperforms, check the list before you rewrite the message.

Diagnosing a weak campaign

When a campaign disappoints, I check the numbers in order, because the order tells you where to look.

Start with acceptance. In HeyReach’s benchmark report, the median campaign landed around 21% acceptance. Below roughly 13% is weak territory, and that usually points to targeting, list precision, or the sender profile. Fix those before you touch the copy. Rewriting a message for a list that never should have been built wastes a week. So go back to your ahrefs or semrush report, and change the competitor, or the page you are competing with. It might be that those backlinks were built with a larger budget than yours, so you are getting answers just about paid placements.

Then check reply rate and acceptance together. The median reply rate was about 22%. If Linkedin reply rate is low but acceptance looks healthy, the problem is almost always the second message. Read it as if you were the recipient. Is it specific to them? Does it give them a reason to respond? Does it ask one clear question?

If both acceptance and reply rate are low, I go back to the list. Targeting problems hide behind weak copy, and the copy gets the blame.

Finally, I compare the reply-to-acceptance conversion rate with the benchmark. The typical campaign converts about 18% of accepted connections into replies. Campaigns under roughly 10% are in the weak range, and that is where I look hardest at the messages sent after acceptance.

The order matters more than the tactics

Earn the connection with a relevant target. Earn the reply with a specific second message. Then scale what works. Doing it in reverse is how you end up with a big list, a full inbox of silence, and a sender account on a warning list.

Email still has its place, and I would not drop it. But for editors, webmasters, and partner marketers who spend their days on LinkedIn, a well-researched message there can get a conversation started that a cold email never would. Test it on a small list, measure the post-acceptance step, and let the numbers tell you where to go next.

FAQ

Is LinkedIn better than email for link building?

Not as a rule. Linkedin feels more personal, and it can build better start of a relationship, but the best is to combine these two in your outreach process.

How do I find editors and webmasters on LinkedIn?

Start with Sales Navigator filters around their beat, then check their recent posts and the sites they link to. Score each target on relevance and recent activity before you add them to a campaign.

How many touches should a link outreach sequence have?

Three works well for me: a connection request, an observation after acceptance, and the ask after a reply. Adding more touches before a reply usually adds annoyance, not results.

Can I automate LinkedIn outreach for partnerships?

You can automate the sending and the follow-up within LinkedIn’s limits. Keep the research and the first observation personal, because automation works best when the message still reads as if a person wrote it.

Is it safe to send link and partnership requests on LinkedIn?

It is safe when you stay within LinkedIn’s limits, avoid scraping and bots, and keep each message relevant to the person. The fastest way to get an account restricted is to send the same pitch to hundreds of people who never asked to hear it.

Related reading

• The Role of Outreach in Broader Search Engine Optimization Plans
• AI-Powered SEO Outreach: Automating Link Building at Scale
• 25 Best Outreach Tools for 2026
• 50+ LinkedIn B2B Statistics for 2026: Users, Ad Costs and ROI

These are the ChatGPT statistics that matter in 2026: users, how people use it, revenue, traffic and what it means for brands that want to show up in its answers. Every number links to its source.

• Top ChatGPT Statistics for 2026
• How Many People Use ChatGPT?
• ChatGPT Usage in the US
• What Do People Use ChatGPT For?
• ChatGPT Traffic and Market Share
• ChatGPT Statistics for Marketers and SEOs
• OpenAI Revenue and Business Statistics
• How Americans Feel About AI Chatbots
• ChatGPT Statistics FAQ
• Related Reading

Top ChatGPT Statistics for 2026

1. OpenAI’s products passed 1 billion weekly active users in August 2026, up from 900 million in February. (Source: AI Magazine)
2. OpenAI reported in October 2026 that more than 1.2 billion people use ChatGPT every week. (Source: AI Magazine)
3. 44% of US adults say they use ChatGPT, up from 18% in 2023. (Source: Pew Research Center, 2026)
4. OpenAI had 50 million paying subscribers by February 2026. (Source: TechCrunch)
5. ChatGPT’s share of generative AI website visits fell from about 76% in June 2025 to about 53% in May 2026. (Source: Similarweb)
6. More than 70% of ChatGPT messages are not related to work. (Source: NBER working paper by OpenAI and Harvard researchers)
7. ChatGPT sends 74.78% of all AI referral traffic to websites. (Source: SE Ranking, 2026)
8. Roughly 26% of ChatGPT responses now contain an ad. (Source: Similarweb)
9. OpenAI has more than two million business customers, and enterprise now brings in most of its revenue. (Source: AI Magazine, citing CFO Sarah Friar via CNBC)
10. Citations appeared in 6.8% of US ChatGPT prompts in May 2026, up from 1.6% a year earlier. (Source: Similarweb)

How Many People Use ChatGPT?

• About one in eight people online uses ChatGPT every week. (Source: AI Magazine)
• By July 2025, ChatGPT had been adopted by around 10% of the world’s adult population. (Source: NBER)
• Early ChatGPT users were mostly male, but the gender gap has narrowed sharply, and growth has been faster in lower-income countries. (Source: NBER)
• OpenAI said January and February 2026 were on track to be its biggest months ever for new subscribers. (Source: TechCrunch)
• Codex, OpenAI’s coding agent, had more than two million weekly users in March 2026. (Source: AI Magazine)

ChatGPT Usage in the US

Pew Research Center has asked US adults about ChatGPT every year since 2023. Its February 2026 survey of 5,119 adults shows steady growth.

• Adults under 50 are about twice as likely as those 50 and older to use ChatGPT (57% vs. 28%). (Source: Pew Research Center)
• 49% of US adults use at least one AI chatbot, up from 33% in 2024. (Source: Pew Research Center)
• 24% of US adults use a chatbot every day, including 12% who use one several times a day and 4% who use one almost constantly. (Source: Pew Research Center)
• ChatGPT (44%) is far ahead of Gemini (24%), Copilot (17%), Meta AI (14%), Grok (8%) and Claude (6%) among US adults. (Source: Pew Research Center)

What Do People Use ChatGPT For?

• Non-work messages grew from 53% to more than 70% of all ChatGPT usage. (Source: NBER)
• “Practical guidance”, “seeking information” and “writing” make up nearly 80% of all ChatGPT conversations. (Source: NBER)
• Writing is the most common work task, while computer programming and self-expression are relatively small shares of use. (Source: NBER)
• 42% of US adults use chatbots to search for information, and 38% of employed adults use them for work tasks. (Source: Pew Research Center)
• 25% use chatbots for fun, 24% to create or edit images or video, 20% for medical advice and 13% to get news. (Source: Pew Research Center)
• 10% of US adults use chatbots for emotional support and 4% for companionship. (Source: Pew Research Center)
• 30% of US adults say chatbots help their productivity, compared with 5% who say they hurt it. (Source: Pew Research Center)
• Similarweb found ChatGPT’s usage mix is the closest of the major assistants to a general search engine, while Claude and Gemini skew more toward programming. (Source: Similarweb)

Because so many people use ChatGPT to look things up, it now competes with Google for the top of the funnel. See our AI search statistics for how this is changing Google itself.

ChatGPT Traffic and Market Share

• Generative AI sites averaged 9.5 billion monthly web visits between June 2025 and May 2026, up 70% year over year. (Source: Similarweb)
• ChatGPT’s share of those visits fell from about 76% to about 53%, while Gemini rose from under 9% to around 27% to 28% and Claude from about 2% to close to 9%. (Source: Similarweb)
• ChatGPT’s absolute visits stayed roughly flat; the category simply grew faster around it. (Source: Similarweb)
• chatgpt.com still gets more web visits than Gemini, Claude and Perplexity combined. (Source: Similarweb)
• ChatGPT’s app has the highest “stickiness” (share of monthly users active daily) of the four big assistants. (Source: Similarweb)
• By year-over-year app growth, ChatGPT is the slowest of the four, behind Claude, Perplexity and Gemini. (Source: Similarweb)

For the challengers, read our Perplexity and Claude statistics. For an earlier rival, see our DeepSeek statistics.

ChatGPT Statistics for Marketers and SEOs

For brands, the question is not how many people use ChatGPT, but whether ChatGPT sends them visits and mentions them in answers.

• ChatGPT sent 74.78% of all AI referral traffic to websites in 2026, down from 79.74% in 2025. (Source: SE Ranking)
• ChatGPT’s referral traffic to websites still grew 27% year over year. (Source: SE Ranking)
• After ChatGPT started showing clickable brand names in May 2026, homepage referrals rose from about 26% to 29% of its referral traffic to about 62% to 63%. (Source: Similarweb)
• Citations appeared in 6.8% of US ChatGPT prompts in May 2026, up from 1.6% in June 2025. (Source: Similarweb)
• Citation rates vary by category: about 23% for travel and hospitality, about 20% for automotive and under 4% for professional services. (Source: Similarweb)
• About a third of ChatGPT ad impressions appear in the first response of a session, and about a fifth appear ten or more messages in. (Source: Similarweb)

In practice, this means a ChatGPT mention often sends people straight to your homepage or makes them search for your brand later. Our guide on how to get cited by ChatGPT and our brand mentions service cover how to earn those mentions.

OpenAI Revenue and Business Statistics

• OpenAI raised $110 billion in February 2026 at a $730 billion pre-money valuation, with $50 billion from Amazon and $30 billion each from Nvidia and SoftBank. (Source: TechCrunch)
• The round closed at $122 billion in March 2026, valuing OpenAI at $852 billion post-money. (Source: AI Magazine)
• OpenAI was earning about $2 billion a month by March 2026. (Source: AI Magazine)
• The Financial Times reported annualised revenue close to $50 billion at the end of September 2026. (Source: AI Magazine)
• OpenAI’s APIs process more than 15 billion tokens a minute. (Source: AI Magazine)
• OpenAI projects revenue growing from $36 billion in 2026 to $350 billion in 2030, according to the FT. (Source: AI Magazine)

Revenue figures for OpenAI come from investor documents and press reports, not audited accounts, and different outlets have published different numbers. Treat them as estimates.

How Americans Feel About AI Chatbots

• 63% of US adults think AI is advancing too quickly; only 2% say too slowly. (Source: Pew Research Center)
• 71% think AI will make their personal information less secure. (Source: Pew Research Center)
• 40% expect AI to have a negative impact on society over the next 20 years, vs. 16% who expect a positive one. (Source: Pew Research Center)

For broader numbers on chatbots and AI tools, see our chatbot statistics and generative AI statistics.

ChatGPT Statistics FAQ

How many people use ChatGPT in 2026?

OpenAI’s products passed 1 billion weekly active users in August 2026, and OpenAI reported in October that more than 1.2 billion people use ChatGPT each week. In February 2026 the figure was 900 million weekly users with 50 million paying subscribers.

What percentage of Americans use ChatGPT?

44% of US adults said they use ChatGPT in Pew Research Center’s February 2026 survey, up from 34% in 2025 and 18% in 2023. Adults under 50 are about twice as likely to use it as adults 50 and older (57% vs. 28%).

Is ChatGPT losing market share?

Yes, in share, not in absolute use. Similarweb data shows ChatGPT’s share of generative AI web visits fell from about 76% in June 2025 to about 53% in May 2026, while its total visits stayed roughly flat. Gemini and Claude took most of that share.

Does ChatGPT send traffic to websites?

Yes. ChatGPT sends about three-quarters of all AI referral traffic, and that traffic grew 27% in 2026. Still, AI platforms as a whole send only 0.32% of all website traffic, so Google remains far bigger as a traffic source.

Related Reading

• AI search statistics
• Zero-click search statistics
• Best AI search engines like Perplexity
• How to optimize brand visibility for AI chatbots
• AI search visibility service