How to Get Cited by Claude: A Small-Site Guide to Claude's Web Search
Claude cites pages it retrieves during a web search, and Anthropic says every search answer includes citations. To be one of them, three things have to be true: Anthropic's search and user bots (Claude-SearchBot and Claude-User) can read your pages, your pages show up in the search results Claude gets, and the page has a short, quotable answer to the question. There's no form to submit and no paid placement. Below is what Anthropic documents, what is only known from outside observers, and a simple way to test your own site.

How Claude decides to search and cite
Claude doesn't search for every question. Anthropic's web search tool documentation says Claude searches when a request depends on information that is current or outside its training data: recent news, prices, statistics, and facts about "specific organizations, people, or products that might have changed." For stable knowledge, like math or general concepts, it answers directly.
When it does search, the same page says "citations are always enabled for web search." Each citation carries the page URL, its title and up to 150 characters of cited text. Search results also include a page_age field, meaning when the page was last updated. In the Claude apps, the help article on web search puts it plainly: "Every response includes citations, so you can easily verify sources yourself."
Two practical takeaways for a small site:
- Target questions that trigger a search. "What is SEO?" probably won't. "Best free rank tracker in 2026" or "does X plugin work with Y" probably will.
- Write sentences that work as a 150-character quote. If your answer only makes sense after three paragraphs of setup, it's hard to cite.
Anthropic's three bots, and which ones matter for citations
Anthropic's crawler help page lists three user agents, each controlled separately in robots.txt:
| User agent | What it does | If you block it |
|---|---|---|
| ClaudeBot | Collects pages that may be used for model training | Your future pages "should be excluded" from training data |
| Claude-SearchBot | Crawls to "improve search result quality" | Anthropic can't index your content for search |
| Claude-User | Fetches a page when a Claude user's question needs it | Claude can't retrieve your page in response to a user query |
For citations, the two that matter are Claude-SearchBot and Claude-User. ClaudeBot is about training, so you can block it and still be citable. If you want that split, a robots.txt like this does it:
User-agent: ClaudeBot
Disallow: /
User-agent: Claude-SearchBot
Allow: /
User-agent: Claude-User
Allow: /
The help page also notes that Anthropic supports the non-standard Crawl-delay directive, that blocking by IP address isn't a reliable opt-out, and that rules need to go in the robots.txt of each subdomain. A common accident is a blanket User-agent: * / Disallow: / left over from staging, or a security plugin that blocks "AI bots" as a group. Our guide to blocking AI crawlers with robots.txt shows how group matching works if you want to allow some bots and not others.
Where Claude's search results come from
Anthropic hasn't published a full list of the search providers behind Claude's web search. What is public:
- In March 2025, Anthropic added Brave Search to its subprocessor list, as TechCrunch reported. Simon Willison compared Claude's citations with Brave's results for the same query and found they matched. That's outside evidence, not an official statement, and it dates from 2025.
- Anthropic now also runs Claude-SearchBot, which by its own description indexes pages for search. So Anthropic's own crawling plays some part as well.
The Brave side has one consequence worth knowing. Brave's crawler page says its bot doesn't use a separate user agent and that "if a domain or page is not crawlable by Googlebot, then Brave Search's bot will not crawl it either." So blocking Googlebot can also keep you out of Brave's index. To check, search site:yourdomain.com on search.brave.com. If nothing shows up, fix basic indexing first; our indexing guide for AI search covers the same groundwork for Bing and ChatGPT.
Write pages Claude can quote
None of this is an Anthropic ranking rule. It follows from how citations work: a URL, a title and a short quoted span.
- Answer first. Put a one- or two-sentence answer right under the heading that asks the question.
- Use question headings people actually type. "How much does X cost?" beats "Pricing considerations."
- Show dates. Results carry a page age, and Claude searches mainly for current information. Write "checked October 2026" next to prices and limits, and update the page when they change.
- Be specific and sourced. A sentence with a number and a link to the primary source is easier to trust and quote than a general claim.
- Make the title say what's on the page. The title travels with the citation, so it should still make sense out of context.
The same habits help with other assistants too; compare our guides on getting cited by ChatGPT and ranking in Perplexity.
Test it: a 20-minute citation check (illustrative)
Illustrative example. Say you run a small blog about home espresso and have a post comparing two grinders.
- Write down 10 questions your posts answer, phrased the way a buyer would ask them ("Is grinder A worth it over grinder B in 2026?").
- Ask each one in Claude with web search on. Note in a sheet: did Claude search, which URLs it cited, and whether yours was one of them.
- Before: your page is never cited. You check robots.txt and find a security plugin blocking "AI crawlers," including Claude-User.
- After: you allow Claude-SearchBot and Claude-User, move the verdict sentence to the top of the post and add the date you tested. Re-run the same 10 questions a few weeks later and compare.
Answers vary from run to run, so treat this as a direction check, not a ranking. Record the date of every run.
What doesn't help
- Paying for placement. There's no ad or submission program for Claude citations.
- Counting on llms.txt. Anthropic's crawler help page doesn't mention it as a signal for Claude's search. It's cheap to add (see our llms.txt guide), but don't expect it to replace crawler access.
- Thin pages written for bots. A page with nothing worth quoting won't get quoted, however open your robots.txt is.
Check your access in 30 seconds
Before rewriting anything, confirm the basics. Satelo's free AI crawler checker reads your robots.txt the way crawlers do and shows, for 12 AI bots including ClaudeBot, Claude-SearchBot and Claude-User, whether a path is allowed or blocked and which line decides it. It also checks for /llms.txt. No signup, and the result is ready in seconds.
FAQ
Does Claude always cite its sources?
When it uses web search, yes. Anthropic's API docs say citations are always enabled for web search, and the Claude help center says every web search response includes citations. Answers from training data alone don't come with links.
Which Anthropic bot do I need to allow to be cited?
Claude-SearchBot, which indexes pages for search, and Claude-User, which fetches pages when a user's question needs them. ClaudeBot is the training crawler and can be blocked without affecting those two.
If I block ClaudeBot, will Claude stop citing me?
Not according to Anthropic's crawler help page, which describes the three bots separately: blocking ClaudeBot excludes your future pages from training data, while search and user retrieval are controlled by Claude-SearchBot and Claude-User.
Does Claude use Google or Brave for search?
Anthropic hasn't published a full list. In March 2025 it added Brave Search to its subprocessor list, and an independent test found Claude's citations matched Brave results. Anthropic also runs its own Claude-SearchBot crawler.
Can I submit my site to Claude?
No. There's no submission form or webmaster console for Claude. Make sure the Anthropic bots aren't blocked, your pages are indexed in regular search engines, and the content answers real questions.
Does llms.txt help me get cited by Claude?
Anthropic's crawler documentation doesn't list it as a signal. It does no harm and takes minutes to create, but crawler access and quotable content matter more.
How can I tell if Claude visited my site?
Look for the ClaudeBot, Claude-SearchBot and Claude-User user agents in your server logs. Anthropic publishes its crawler IP list at claude.com/crawling/bots.json so you can verify the requests are genuine.
Why does Claude cite me for one question but not a similar one?
Claude decides per request whether to search, and results change over time. Track a fixed set of questions with dates in a sheet and look at trends over several runs, not single answers.
Read next




Check your site's Quality Score
Free, no sign-up: a 0–100 score from mostly open data. Then add your site to Satelo and get partner articles from sites of your level.
Enter a domain like example.com
Add my site