When a buyer asks Claude to compare vendors in your category, Claude builds that answer mostly out of company websites.
Not review platforms. Not community threads. Company and product domains account for 64% of everything Claude cites, according to an OtterlyAI analysis of 379,321 Claude citations across 16,406 unique domains in June 2026.
That one number reorders the priority list for anyone trying to appear in Claude's answers, and it points somewhere unusually convenient: the part of your AI visibility you already own outright.
The catch is that very few AEO programs are built around it, and a surprising number of brands have never checked whether Claude is able to read their site at all.
Why Claude matters now
Claude spent two years as the platform marketing teams knew about and did not track. That window has closed.
The consumer numbers moved fast. Sensor Tower's State of AI 2026 report put Claude at 245 million monthly users and 10.3% of the global AI assistant app market as of May 2026, the same month ChatGPT's share fell below 50% for the first time since launch. Similarweb's worldwide panel shows Claude's share of generative AI web traffic rising from 1.6% to 8.9% over twelve months, the largest proportional gain of any platform tracked.
The B2B numbers moved faster. This is the part that matters for anyone selling to businesses. Goodie's Wave 2 report pulled GA4 referrer data across a panel of 41 brand sites and found that Claude's share of measurable B2B AI referrals went from 1.4% to 18.5% in eight months. Over the same period ChatGPT's share of that pie fell from 89.1% to 63%.
Put those two datasets side by side and the skew is the story. Claude holds roughly 9% of worldwide AI assistant web visits, and 13.4% in the US per Similarweb, but takes about 18.5% of measurable B2B AI referrals. Depending on which baseline you pick, that is somewhere between a 1.4x and 2x over-index in the business category. The denominators are not identical and the B2B panel is small, so treat it as a direction rather than a coefficient. The direction is consistent: Claude punches above its consumer weight in exactly the category where deals get made.
The reason is audience composition. Anthropic leads enterprise LLM API spend at 40%, ahead of OpenAI at 27% and Google at 21%, and Claude's consumer usage skews toward research, comparison, technical evaluation, and professional work. Those are the queries that precede a purchase decision. A brand that is invisible in Claude is invisible at the moment a buyer is building a shortlist.
What Claude actually cites
The OtterlyAI study sorted all 379,321 citations into 11 source categories. It is the most substantial public dataset on Claude's sourcing behavior, and the full distribution is worth sitting with, because the shape of it is not what the standard AEO playbook predicts.
| Source category | Share of citations |
|---|---|
| Brand (company and product sites) | 64.0% |
| News and media | 14.9% |
| Blogs and personal sites | 5.4% |
| Education | 2.7% |
| Government and NGO | 2.4% |
| Encyclopedia | 2.1% |
| Community and forum | 1.3% |
| Social media | 0.9% |
| Video | 0.3% |
Nine of the eleven categories shown; the remainder sits in minor buckets and rounding.
Three findings deserve attention.
Brand domains are the backbone, and no brand owns them
First-party company and product sites account for 64% of everything Claude cites, more than four times the next category. But that dominance is spread thin rather than concentrated. The brand category spans 10,842 distinct domains, and the single most-cited brand domain accounts for under 3% of brand citations.
There is no incumbent to unseat. The field is contestable.
Reddit was cited zero times
Across all 379,321 citations, social platforms accounted for 0.9%. Within that sliver, LinkedIn was 43.1%, and Reddit appeared zero times. X showed up 56 times, about 1.6% of an already negligible slice.
In most published citation studies, Reddit ranks at or near the top of the most-cited domains on ChatGPT and Perplexity. The most common piece of AEO advice in circulation right now is some version of "build community presence and get discussed on Reddit." For Claude, that advice does nothing.
The little social content Claude does cite skews professional and document-shaped. LinkedIn, ResearchGate, and SlideShare appear well ahead of Instagram and TikTok relative to their size. Long-form and document content travels into Claude's answers. Feed content does not.
The tail is unusually long
The top 10 domains account for just 9.5% of all citations. Reaching 60% of citations requires roughly 500 domains, and 32.9% of cited domains were cited exactly once.
Traditional search concentrates. Claude distributes. For a mid-sized brand with modest domain authority, that distribution is the single most encouraging finding in the dataset: you do not need to be a top-10 domain to be cited. You need a page that answers the specific question cleanly.
One more useful property: the pattern looks stable. Comparing the first five days of the study window against the last five, no category moved by more than about one percentage point. That is a short window and not proof of durability, but nothing in the data suggests the mix is churning week to week.
Why Claude behaves so differently
It is worth putting a number on how different. When OtterlyAI compared Claude's cited sources against ChatGPT's over the same period, only 13% of cited domains overlapped, and just 4.2% at the URL level. The two assistants that send the most B2B referrals agree on roughly one source in eight.
That divergence is architectural, not accidental.
ChatGPT assembles answers by consensus. It reaches for what the open web already agrees on, which is why community platforms, encyclopedias, and review aggregators dominate its citation pool. Claude synthesizes, and reaches for primary material it can read and verify directly.
Two supporting patterns show up in independent analyses. Oltre's study of 2,170 Claude-cited URLs found that only 3% were domain homepages while 56% sat under a /blog/ path, and that mainstream news outlets and large social platforms were essentially absent from the sample. Separately, Muck Rack's media analysis found that Claude's top 100 most-cited outlets carry roughly 50% lower average monthly visitors than ChatGPT's top 100, meaning Claude reaches for niche and trade publications where ChatGPT reaches for scale.
The through-line: Claude selects for reference value, not popularity. Specific pages, not front doors. Verifiable claims, not marketing language.
Your own website is the battlefield
For most engines, the uncomfortable truth of AEO is that your site is a minority of the citation pool and the real work is off-site. Claude inverts that. At 64%, first-party content is the primary route to a Claude citation, and it is the one part of the system you fully control.
That changes what "good" looks like on your own domain.
Build reference pages, not brochures. Claude cites content that resolves a question. Product, solution, comparison, integration, and "how it works" pages should read like documentation: clear definitions, explicit scope, stated limitations, supporting evidence with dates. Pure marketing copy performs worse than neutral technical writing.
Write for passage-level extraction. Citations happen at the section level, not the page level. A single well-structured paragraph can earn a citation from an otherwise ordinary page, and an excellent page with tangled structure can be read and never cited. Each H2 section should be a self-contained, quotable unit that opens with a direct answer.
Compete on breadth. Because a third of cited domains appear only once, coverage creates surface area. A library of focused pages, each resolving one specific question, matches more real prompts than a single comprehensive guide.
Date everything and cite your sources. Claude weights verifiable, current claims. A sentence referencing a named study with a date carries more weight than one referencing "recent research."
The technical layer most teams have wrong
This is the part worth auditing this week, because a meaningful number of brands are blocking themselves without knowing it.
Anthropic operates three separate crawlers, each with its own robots.txt user agent and its own consequence when blocked:
- ClaudeBot collects public web content that may be used to train future models.
- Claude-SearchBot indexes content to support Claude's search results. This is the visibility gate.
- Claude-User fetches a specific page when a Claude user asks a question that requires it.
The critical detail is that these operate independently. Blocking ClaudeBot does not block Claude-SearchBot or Claude-User, and blocking one does not imply the others. Anthropic's own documentation states that disabling Claude-SearchBot prevents its system from indexing your content, which may reduce your site's visibility and accuracy in user search results.
Here is the trap. During the 2024 wave of AI training opt-outs, a great many sites added a blanket User-agent: ClaudeBot / Disallow: / rule and never revisited it. On its own, that rule is a defensible policy choice, and it does not cost you citations. The damage comes from the broader version: teams that pasted a catch-all AI-blocking template caught Claude-SearchBot and Claude-User in the same net and removed themselves from Claude's answers entirely, as a side effect of a training decision they thought they were making.
Two more technical checks worth running:
Render server-side. Anthropic does not publish a rendering guarantee for its crawlers, so content that only appears after client-side hydration should be assumed unread. Server-side rendering removes the question.
Check every subdomain. Robots rules are host-specific. If your marketing site allows Claude's crawlers and your docs subdomain does not, you have blocked the most citable content you own.
Where Claude fits in an AEO program
Claude belongs in your tracked lineup, and Sentient AEO's Visibility Tracking covers it alongside ChatGPT, Gemini, Perplexity, and Google AI surfaces. But it needs honest scoping, and there are two places where the industry data does not support confident claims.
The conversion numbers disagree with each other. Published estimates for Claude referral conversion range from about 5% in the Seer and Position Digital breakdown to 16.8% in Exposure Ninja's analysis. Both are well above Google organic, but a three-fold spread between studies means neither should be presented to a board as a planning number. Use the direction, not the decimal.
Referral tracking understates Claude badly. Anthropic's crawlers fetch a very large number of pages relative to the visits they return, with Nobori's Q2 2026 analysis putting the ratio in the thousands of pages per referral against roughly five for Google. Agentic and in-conversation reading leaves no session in GA4. If you judge Claude by referral volume alone, you will conclude it does not matter, because referrals are not the primary thing it produces.
The right measures are share of answer, citation share against named competitors, and how Claude describes you when it does mention you.
And the work mostly transfers. Reference-grade first-party content, clean heading structure, dated and sourced claims, and crawler access are not Claude-specific tactics. They improve extraction on every engine. What is Claude-specific is the weighting: the off-site community layer that drives ChatGPT visibility earns you close to nothing here, and the on-site layer that most brands treat as a checkbox is where two-thirds of the citations come from.
The mistake is not failing to build a separate Claude program. The mistake is assuming that one AEO scorecard describes both engines, when the data says they agree on one source in eight.
Sentient AEO helps brands build and measure AI search visibility across ChatGPT, Google AI Overviews, and Perplexity, with additional coverage for Gemini, Google AI Mode, Grok, and Claude. If you're trying to understand where your brand stands in the AI answer layer, get in touch with us for an AEO audit: [email protected]
Citations
-
Brand and product domains at 64% of 379,321 Claude citations across 16,406 unique domains, June 2026, with the full 11-category distribution — OtterlyAI, Claude Citation Study Reveals What Actually Gets Cited: https://otterly.ai/blog/claude-ai-citation-study/
-
Claude at 245 million monthly users and 10.3% of the global AI assistant app market in May 2026, the month ChatGPT fell below 50%, per Sensor Tower's State of AI 2026 — TechCrunch, ChatGPT's market share slips below 50% for first time: https://techcrunch.com/2026/06/16/chatgpts-market-share-slips-below-50-for-first-time/
-
Claude's share of generative AI web traffic rising from 1.6% to 8.9% over twelve months, per Similarweb — PPC Land, ChatGPT drops to 52.7% as Claude triples its AI traffic share: https://ppc.land/chatgpt-drops-to-52-7-as-claude-triples-its-ai-traffic-share/
-
Claude's share of measurable B2B AI referrals rising from 1.4% to 18.5% in eight months while ChatGPT's fell from 89.1% to 63%, across a 41-site GA4 panel — Goodie, AI Search Traffic Report 2026 (Wave 2): https://higoodie.com/blog/ai-search-traffic-report-2026/
-
Claude at 13.4% of US AI assistant web visits per Similarweb May 2026 data, and Anthropic leading enterprise LLM API spend at 40% ahead of OpenAI (27%) and Google (21%) per Menlo Ventures — Momentic, Top Generative AI Chatbots & LLMs by Market Share: https://momenticmarketing.com/blog/top-ai-chatbots
-
The brand category spanning 10,842 distinct domains with the most-cited brand domain under 3% of brand citations — OtterlyAI, Claude Citation Study Reveals What Actually Gets Cited: https://otterly.ai/blog/claude-ai-citation-study/
-
Social media at 0.9% of citations, LinkedIn at 43.1% of that slice, Reddit cited zero times, X cited 56 times — OtterlyAI, Claude Citation Study Reveals What Actually Gets Cited: https://otterly.ai/blog/claude-ai-citation-study/
-
Top 10 domains at 9.5% of citations, roughly 500 domains needed to reach 60%, 32.9% of domains cited exactly once, and category shares moving by no more than about one point between the first and last five days of the window — OtterlyAI, Claude Citation Study Reveals What Actually Gets Cited: https://otterly.ai/blog/claude-ai-citation-study/
-
13% domain-level and 4.2% URL-level overlap between Claude's and ChatGPT's cited sources over the same period — OtterlyAI, Claude Citation Study Reveals What Actually Gets Cited: https://otterly.ai/blog/claude-ai-citation-study/
-
Only 3% of 2,170 Claude-cited URLs being domain homepages, 56% sitting under a /blog/ path, and mainstream news and large social platforms essentially absent — Oltre, How Claude Picks Sources: A Technical Breakdown: https://www.oltre.ai/blog/how-claude-picks-sources-technical-breakdown-claude-citations/
-
Claude's top 100 most-cited outlets carrying roughly 50% lower average monthly visitors than ChatGPT's top 100 — Muck Rack, 3 new insights on how Anthropic's Claude cites media: https://muckrack.com/blog/how-claude-cites-media
-
ClaudeBot, Claude-SearchBot, and Claude-User as three separate crawlers, and the statement that disabling Claude-SearchBot may reduce a site's visibility and accuracy in user search results — Anthropic Privacy Center, Does Anthropic crawl data from the web, and how can site owners block the crawler?: https://privacy.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler
-
Blocking ClaudeBot not blocking Claude-SearchBot or Claude-User, and each bot honoring only its own robots.txt rule — Search Engine Land, Anthropic clarifies how Claude bots crawl sites and how to block them: https://searchengineland.com/anthropic-claude-bots-470171
-
Claude's crawlers operating independently, making robots.txt decisions more granular — Search Engine Journal, Anthropic's Claude Bots Make Robots.txt Decisions More Granular: https://www.searchenginejournal.com/anthropics-claude-bots-make-robots-txt-decisions-more-granular/568253/
-
Claude referral conversion at roughly 5%, per Seer Interactive and Position Digital data — The SEO Autopilot, AI search referral conversion vs Google: https://the-seo-autopilot.com/en/stats/ai-vs-google-referral-conversion
-
Claude referral conversion at 16.8%, per Exposure Ninja — Yahoo Finance, 73% of B2B Buyers Use AI Tools in Purchase Research: https://finance.yahoo.com/sectors/technology/articles/73-b2b-buyers-ai-tools-231200431.html
-
Anthropic's crawl-to-refer ratio in the thousands of pages per referral against roughly five for Google, Q2 2026 — Nobori, Crawl-to-Refer Ratio 2026: AI Bots Take 4,580, Give 1: https://nobori.ai/blog/crawl-to-refer-ratio-ai-crawler-traffic-b2b-2026
