Claude does not rank pages. It evaluates whether your content is worth quoting. That is a different problem to solve, and it is why most businesses built to sell are invisible on Anthropic's platform.
Why does Claude ignore most business websites?
Your website was built to sell. Claude was built to evaluate. That mismatch is why most businesses are invisible on Anthropic's platform.
Claude's selection model is editorial. It looks for specific claims, named sources, and transparent methodology. Content that reads as marketing copy gets passed over. No exceptions.
Web search launched in March 2025. There is no toggle. When the model determines current information would improve an answer, it searches automatically, retrieves sources, evaluates them, and cites the ones it considers substantive. Your business is either in that list or it is not.
This is where Claude separates from every other platform. ChatGPT leans on Bing's index. Perplexity rewards earned media and community presence. Generative engine optimisation covers the shared principles, but each platform weights them differently. Claude is closer to peer review than to search ranking. A page that says "we tested this across 200 companies" is citable. One that says "our proven approach delivers results" is not.
80% of URLs cited by AI assistants do not rank in Google's top 100. Traditional rankings are an unreliable predictor of whether Claude will cite you. The businesses appearing in Claude's answers are not there because they rank well on Google. They are there because their content reads like a primary source.
Source: Ahrefs, 2025
What does Claude look for in a source?
Specificity is the price of admission. A page with verifiable claims, named data, and described methodology gets treated as a primary source. Generic assertions get ignored. If someone could verify it, Claude cites it. If not, Claude moves on.
Five signals carry the most weight:
- →Methodology transparencySpecific claims backed by named data. "We audited 50 B2B companies and most appeared on zero AI platforms for their core buyer queries" is citable. "Our proven approach delivers results" is not.
- →Content structureQuestion-based headers that mirror natural-language queries. Self-contained sections that Claude can extract without needing the rest of the page. Logical H2/H3 hierarchy.
- →Schema as trust signalFAQPage schema gives Claude pre-labelled question-and-answer pairs. Service and BreadcrumbList schema provide navigation context. Schema does not drive citations alone, but it improves extraction accuracy once content is selected.
- →Brand mentions over backlinksWeb mentions correlate 0.664 with AI visibility versus 0.218 for backlinks (Ahrefs, 75,000 brands, 2025). Claude does not care how many sites link to you. It cares how many credible sources mention you in the context of your expertise.
- →Author credentialsNamed author with real credentials. Content attributed to a specific expert carries more weight than anonymous or "Staff Writer" content.
None of this is about gaming a system. Claude rewards the same qualities that make content useful to a human reader. Be specific. Show your evidence. Name your sources.
How is Claude's architecture different from ChatGPT and Perplexity?
Most companies block all AI crawlers and wonder why they are invisible. Claude operates three separate bots: ClaudeBot for training data, Claude-SearchBot for search indexing, and Claude-User for retrieval. Blocking ClaudeBot prevents training data collection but does not prevent search indexing. The distinction matters.
The architecture differs across platforms in ways that change how content is selected. Claude reaches the web through search tools on demand, with no known publisher partnerships, and selects on editorial quality and substance. ChatGPT runs primarily on Bing plus its own crawler, carries data deals with publishers such as Reuters, FT, News Corp and Reddit, and weights referring domain count and Bing correlation. Perplexity runs its own 200-billion-URL index with Comet Plus revenue sharing, and weights freshness (around 40%) and earned media. The search behaviour differs too: Claude triggers a web search only when a query needs it, ChatGPT browses on an always-on basis, and Perplexity crawls continuously in real time.
Anthropic has not disclosed what Claude was trained on. There are no known publisher deals. Content earns its way in on substance, not through a licensing agreement.
Claude's Citations API (launched June 2025) references exact sentences and passages within a page. This is why structural work matters. Claude is not just deciding whether to cite your page. It is deciding which specific passage to quote.
Only 11% of cited domains appear across multiple AI platforms. Visibility on ChatGPT does not transfer to Claude. Platform-specific optimisation is not optional.
Source: Profound, 680M citations, 2025
What does Claude optimisation actually involve?
Google rewards ranking signals. Claude rewards editorial quality. That distinction changes what optimisation actually means. The index is different, the signals carry different weight, and the technical requirements go beyond what most agencies test for.
Three areas, each specific to how Claude selects sources:
- 01Content substanceClaude rewards methodology transparency, primary-source data, and well-attributed claims. Marketing copy gets ignored. Show your working: "We audited 200 sites and found X" is citable. "Our proven approach" is not. 80% of URLs Claude cites do not rank in Google's top 100 (Ahrefs, 2025). Substance is the selection filter.
- 02Authority through mentions, not linksBrand mentions correlate 3x stronger than backlinks for AI citation (0.664 vs 0.218, Ahrefs, 75,000 brands). Claude does not count who links to you. It reads whether credible sources mention you in the context of your expertise. Get named in industry publications, analyst coverage, and third-party research Claude already trusts, and make sure your entity SEO clarifies who you are so Claude connects those mentions to the right business.
- 03Technical clarityClean HTML, proper schema, server-side rendering. Claude operates three bots: ClaudeBot (training), Claude-SearchBot (web search), and Claude-User (retrieval). Each needs to parse your content without JavaScript rendering. Block ClaudeBot if you choose, but leave Claude-SearchBot open or you vanish from search entirely. A technical AI visibility review finds the crawler-access, rendering, and schema-delivery blockers stopping these bots from reading you.
Claude's Citations API (launched June 2025) references exact sentences and passages. Content structured in self-contained sections benefits directly. For the broader picture across all six platforms, see our AI search visibility services.
What happens in a Claude visibility audit?
Most businesses have no idea how Claude sees them. An AI visibility audit answers that question. Real buyer queries are run through Claude in your market, every response recorded, and your citation strength scored against competitors. That is one layer of a full assessment that reads your site across nine dimensions of commercial AI search visibility health, scored Strong, Mixed, or Weak on a live call.
The signals it reads include technical foundation, content extractability, AI crawler access, schema markup, brand entity coherence, competitive positioning, and content quality. Each recommendation is tied to the Claude-specific signal it addresses.
A financial services firm in London had strong domain authority but generic service descriptions. Claude cited them on one of twelve queries. Their competitor, a smaller firm with detailed methodology pages and named author credentials, appeared on nine. The gap was not authority. It was substance.
Claude visitors convert at 5%, nearly 3x the rate of organic search (1.76%, Semrush, 2025). Generative AI tools are now the single most cited meaningful interaction for B2B purchase research, ahead of vendor websites and sales conversations (Forrester, Buyers' Journey Survey 2025, 18,000 global buyers). The buyers Claude sends are not browsing. They have already had their question answered. They come to confirm and contact.
Source: Semrush, 2025; Forrester, 2025
What to do next
Run the buyer-intent queries relevant to your market through Claude and record whether you appear. If you do, note whether Claude treats your content as a primary source or passes over it for a competitor with more substantive pages.
The businesses building substance now will own the citations as Claude's user base grows. The ones that start later will be paying to displace them. If the audit shows your pages, page roles, schema, and entity structure are not clear enough for Claude to extract, a Visibility Rebuild fixes that substance and structure directly. For a full assessment across all six AI platforms, start a free AI visibility audit.
Mo Walji founded 1DOT in 2015. The company now helps B2B companies become visible to AI search platforms across ChatGPT, Perplexity, Gemini, and Google AI Overviews. Mo works directly with construction, manufacturing, professional services, and education companies in the UK and US.
Read more about Mo →