Claude SEO: how Claude finds, reads and cites web pages

That matters because people now ask Claude the questions they used to type into a search box, and a page Claude can't reach can't be cited. This guide covers when Claude searches, what Anthropic says about its search index, its three crawlers, what you can and can't control, how to measure Claude visibility, and what nobody can promise.

Does Claude search the web?

Yes. Claude can search the web in the Claude apps and through Anthropic's API. Anthropic's help article says that when you ask about topics that benefit from current information, "Claude invokes a search tool to inform and ground its generated responses with content from the live web. Every response includes citations, so you can easily verify sources yourself" (Claude Help Center).

How it's switched on depends on the account, as of October 2026:

• Most accounts turn on "Web search" from the + menu in a chat.
• Team and Enterprise: "An Owner or Primary Owner must first enable web search for the entire workspace in Organization settings > Capabilities."
• The new Claude experience: "there's no web search toggle. Claude searches the web when it helps."

Claude doesn't search for everything. Anthropic's developer docs for the web search tool say "Claude determines when to search based on the prompt," and that it searches when "the request depends on information that is current, changing, or outside its training data": recent news, current prices, and "information about specific organizations, people, or products that might have changed" (Claude API docs). It answers stable-knowledge questions, such as math or coding concepts, without searching.

For SEO, the takeaway is that pricing pages, product pages, release notes and anything about your company are exactly the kinds of pages Claude looks up live. Older facts about your brand may instead come from training data, which you can't edit after the fact.

Where do Claude's answers come from?

Claude's answers come from two places: what the model learned in training, and what a web search returns at the moment someone asks. Only the second one produces cited links to your page.

For the search side, here is what Anthropic documents, and only that:

What Anthropic doesn't publish is how these fit together: which provider handles which query, how much comes from Claude-SearchBot's own crawling, or how results are ranked before Claude reads them. Treat any article that states a single search engine "behind" Claude as a guess unless it links to Anthropic saying so. The Bing statement above covers image results only.

Each search result Claude receives carries a page_age field that Anthropic's API docs describe as "When the site was last updated" (Claude API docs). Anthropic doesn't say how Claude weighs it, but it means the date your page shows is visible to the model.

Which crawlers does Claude use?

Claude relies on three Anthropic bots, each with its own robots.txt token, and they do different jobs. Anthropic's crawler article describes them, and what blocking each one does (Anthropic Help Center):

All three honor robots.txt; Anthropic says its bots "respect 'do not crawl' signals by honoring industry standard directives in robots.txt." So a site can opt out of training and still be findable in Claude:

User-agent: ClaudeBot
Disallow: /

User-agent: Claude-SearchBot
User-agent: Claude-User
Allow: /

If you already have a User-agent: * group with Disallow lines, repeat them under the Claude group, because a named group replaces the * group for that bot. The full detail, including the user-agent string, Crawl-delay and how to verify the bot by IP, is in our ClaudeBot guide. The AI crawlers guide covers the equivalent bots from OpenAI and others, and editing robots.txt on WordPress shows where the file lives on that platform.

Can you submit pages to Claude?

No. As of October 2026, Anthropic documents no way to submit a URL to Claude or its search index. Its crawler article offers robots.txt controls, a published IP list and a contact address (claudebot@anthropic.com), and nothing like Google's URL Inspection tool or the IndexNow protocol (an open "this page changed" notice that Bing and other participating engines accept; Google doesn't use it).

The levers you do have are indirect:

1. Let Claude-SearchBot and Claude-User in. Check robots.txt, and check that your firewall or bot-protection service isn't challenging them. Anthropic says its bots "will not attempt to bypass CAPTCHAs."
2. Put the content in the HTML. Anthropic's developer docs say its web fetch tool "currently does not support websites dynamically rendered with JavaScript" (Claude API docs). That page documents the API tool, not every Claude product, but server-rendered text is the safe choice everywhere.
3. Be findable on the open web. Anthropic runs its own search crawler and also lists outside web search providers, and it doesn't say which one answered a given question. Being crawlable by Claude-SearchBot is the part you control; Anthropic doesn't say whether being in any other index helps. See how AI search chooses its sources for how retrieval works across engines.

If a page isn't showing up anywhere, Google included, start with the basics: run a free check of which pages Google has indexed, then fix what isn't.

What can you control to get cited by Claude?

You control access, clarity and accuracy; Claude controls the choice. The same practices that help with any AI answer engine apply, and Anthropic's docs point to a few Claude-specific details.

• Answer the question near the top. When Claude cites a source through the API, each citation includes "Up to 150 characters of the cited content" (Claude API docs). Short, self-contained sentences that state a fact are easy to quote. Our AEO guide (answer engine optimization: making pages easy for AI answer engines to find, trust and quote) covers how to write them.
• Keep facts current and dated. Claude searches when information is current or changing, and its results carry a last-updated date. Update pages when facts change, not just the date; see content freshness for AI search.
• Cover your brand's own facts. Since Claude looks up "specific organizations, people, or products that might have changed," make sure your pricing, product and about pages state those facts plainly.
• Earn mentions elsewhere. Anthropic says Claude "processes multiple sources to find relevant content" (Claude Help Center), so third-party pages that describe you accurately give it more to work with.

None of this is special "Claude AI SEO". It's ordinary good practice for LLM SEO, applied to an engine whose search index you can't inspect. For the OpenAI version of the same question, see ChatGPT SEO.

How do you measure your visibility in Claude?

You measure it yourself, because as of October 2026 Anthropic publishes no visibility or citation report for site owners. Its crawler article covers robots.txt and contact details only. There are four practical methods:

1. Ask Claude directly. Run the questions your customers ask, with web search on, and record which sources it cites. Use the same prompts each time and note the date, because answers vary from run to run.
2. Script it through the API. Anthropic's web search tool returns structured citations with a URL and title, and "Citations are always enabled for web search." As of October 2026 it costs "$10 per 1,000 searches, plus standard token costs" (Claude API docs). Its allowed_domains and blocked_domains options (you can use one or the other, not both) let you test how Claude answers when limited to, or kept away from, certain sites.
3. Watch your server logs. Requests from Claude-SearchBot and Claude-User show which pages Anthropic's search and user fetches reach. Verify them against Anthropic's IP list, as the ClaudeBot guide explains.
4. Use a tracking tool. Several AI visibility tools include Claude on some plans; our comparison of AI visibility tools lists which ones, with prices as of October 2026.

If you use Claude for SEO work yourself, FreshlyCited's MCP server (MCP is the Model Context Protocol: a standard way for AI assistants such as Claude or Cursor to use a tool directly) can be added to Claude.ai as a connector or to Claude Code. Claude can then check whether Google has indexed your URLs. It reports Google's index status, not Claude citations. MCP access is on the Agency Pro and Agency Scale plans.

What can nobody promise about Claude?

Nobody can promise that Claude will cite a page, or when. Anthropic decides how Claude searches and which results it uses, and it doesn't publish ranking factors for Claude's search.

Timing studies exist, but they describe averages, not outcomes for your site. In a July 2026 vendor study of 1,240 sources (740 new pages on brands' own sites and 500 earned mentions) across 132 brands, MaxAEO, which sells AI visibility monitoring, reported that "Claude with web access is slowest at a median of 22 days" from publishing a new page on the brand's own site to its first citation, the slowest of the seven AI surfaces it tracked (MaxAEO). That's one vendor's medians, measured February–May 2026 and not independently replicated, and we have no Claude citation data of our own.

What you can control is narrower and still worth doing: let Claude's search bots in, serve real HTML, keep facts current, and, separately, make sure Google has indexed the page for Google Search and its AI Overviews. Our AI search optimization guide covers the rest of the picture.

Check that your pages can be found

Claude makes its own call on what to cite. Claude, ChatGPT, Perplexity and other assistants use their own crawlers and search partners. We don't submit to them, and we don't claim to. What we do is the Google side: check whether Google has indexed each of your pages, ask Google to look at the ones it hasn't, and re-check on day 1, 3 and 7, because Google's AI Overviews and AI Mode only link to indexed pages. For sites you verify with an IndexNow key, we also send the change to Bing, free. The free plan gives you 10 free links to try, then 100 free checks a month, with no card. Check my pages free.