How do I get my site recommended by Claude?

Allow Claude-SearchBot and Claude-User in robots.txt and at your firewall, make sure your key pages say plainly what you offer and for whom, and get mentioned on the sites Claude already cites for your category's questions.

If I block ClaudeBot, will Claude stop citing me?

No. ClaudeBot is Anthropic's training crawler. Claude's search uses Claude-SearchBot (indexing) and Claude-User (live fetches), each with its own robots.txt token.

Does Claude always search the web?

No. Anthropic's docs say Claude searches for things that change, such as recent events, prices, products and organizations, and answers from what it already knows otherwise. For a product question, what Claude learned in training and what it finds by searching both matter.

Which search engine does Claude use?

Anthropic's documentation doesn't name the provider for text web results; it says image results are powered by Bing. Being indexed by the major search engines and crawlable by Claude-SearchBot covers you either way.

I meant using Claude to do SEO work. Is this the right page?

This page is about getting your site cited by Claude. Using Claude to write or research SEO content is a different job; the steps here decide whether Claude can recommend you at all.

Anthropic · AI engine guide

Claude SEO: how to get cited and recommended by Claude

Claude searches the web when a question needs current information, and every web-search answer cites its sources. Two Anthropic crawlers decide whether your site can be part of that: Claude-SearchBot indexes pages for Claude's search, and Claude-User fetches pages when someone asks about them. ClaudeBot is for training only. Allow the first two, make your key pages answer buyer questions plainly, and get mentioned by the sources Claude already cites.

Crawlers: Claude-SearchBot (ai search): indexes pages to improve the quality of Claude's search results
Claude-User (user-triggered fetch): fetches pages when a person asks Claude a question that needs them
ClaudeBot (model training): collects pages that may train future Claude models; it does not affect search
Where results come from: Anthropic's docs describe Claude grounding answers in content from the live web without naming a provider for text results. Its help center says image results are powered by Bing.
How sources show up: Every web-search answer includes citations to its sources, source links and relevant quotes. In the API, citations are always on and include the URL, title and quoted text.
Your controls: Claude-SearchBot in robots.txt: Blocking it stops Anthropic indexing your content for search, which Anthropic says may reduce your visibility in Claude's search results.
Claude-User in robots.txt: Blocking it stops Claude retrieving your pages in response to a person's question.
ClaudeBot in robots.txt: Opts your future pages out of training data. No effect on search.
Crawl-delay: Anthropic supports the non-standard Crawl-delay line for its bots.

Live check

Can Claude's crawlers reach your site?

How GetPicked measures Claude

Our Pick Check asks Claude the questions your buyers ask through the Anthropic Messages API with the web_search tool, via Vercel AI Gateway (model anthropic/claude-haiku-5.5), several times per question, and counts how often the answer names or cites your site. Each rate comes with a likely range, because answers vary from run to run. How we measure.

Run a free audit

How Claude decides what to cite

Claude doesn't search for everything. Anthropic's documentation says it searches when an answer depends on something that changes (recent events, current prices, products, organizations, people) and answers from what it already knows when it doesn't. For "what's the best tool for X", that usually means a search, and the answer cites what it found: Anthropic says every web-search response includes citations, source links and relevant quotes.

So there are two routes into a Claude answer:

Search. Claude-SearchBot indexes pages to improve Claude's results, and Claude-User fetches pages live for a person's question. If either is blocked, you drop out of that route.
What the model already knows. For questions it answers without searching, Claude relies on training. ClaudeBot collects pages for training; allowing it is how future models learn about you, and blocking it is a legitimate choice for publishers.

What to fix first, in order

Access. Run the live check above. Claude-SearchBot and Claude-User should be allowed. Anthropic warns that blocking by IP may not work reliably because its bots then can't read robots.txt, so use robots.txt rules, and make sure a firewall isn't challenging them. Anthropic also says its bots don't bypass CAPTCHAs, so a challenge page is a block.
Every subdomain. Anthropic's rules apply per host: blog.example.com and shop.example.com each need their own robots.txt. Check them all.
Pages that answer the question. Claude cites quotes, so pages with clear, quotable statements win: one paragraph that says what you are, who it's for and what it costs; specific comparison pages; up-to-date pricing.
Mentions on sites Claude cites. Look at which sources Claude cites when it answers your buyers' questions. The Pick Check lists them per engine. Getting listed or reviewed there is often faster than changing your own pages.

Training: allow or block?

ClaudeBot doesn't affect search, so this is a business decision rather than an SEO one. If your value is the text itself (news, research, courses), blocking it is reasonable. If you sell a product, allowing it means future Claude models are more likely to know what you are before they search. Either way, keep the two search crawlers allowed.

What not to do

Don't block "Claude" with a wildcard. A User-agent: * disallow or a CDN AI-bot toggle usually catches all three crawlers.
Don't rely on one conversation to judge progress. Claude's answers vary between runs; measure rates, not anecdotes.
Don't hide text for models. Claude quotes what it cites; make the visible page say the thing.

Measuring progress

GetPicked asks Claude your buyers' questions through Anthropic's API with web search on, several times each, and reports how often Claude names or cites you, with a likely range, and which sources it cited instead.

Questions

How do I get my site recommended by Claude?: Allow Claude-SearchBot and Claude-User in robots.txt and at your firewall, make sure your key pages say plainly what you offer and for whom, and get mentioned on the sites Claude already cites for your category's questions.
If I block ClaudeBot, will Claude stop citing me?: No. ClaudeBot is Anthropic's training crawler. Claude's search uses Claude-SearchBot (indexing) and Claude-User (live fetches), each with its own robots.txt token.
Does Claude always search the web?: No. Anthropic's docs say Claude searches for things that change, such as recent events, prices, products and organizations, and answers from what it already knows otherwise. For a product question, what Claude learned in training and what it finds by searching both matter.
Which search engine does Claude use?: Anthropic's documentation doesn't name the provider for text web results; it says image results are powered by Bing. Being indexed by the major search engines and crawlable by Claude-SearchBot covers you either way.
I meant using Claude to do SEO work. Is this the right page?: This page is about getting your site cited by Claude. Using Claude to write or research SEO content is a different job; the steps here decide whether Claude can recommend you at all.

Sources

Checked against these pages on 2026-10-18. Operators change their crawlers; check the source before relying on a detail.