How to Get Your Brand Cited by Claude: What Anthropic's AI Actually Pulls

To get cited by Claude, you need three things at once: your pages must be technically reachable by Anthropic's crawlers (ClaudeBot, Claude-User, and Claude-SearchBot), your facts must be repeated consistently across the third-party sources Claude already trusts, and each page must lead with a direct, extractable answer under a named author. Claude does not rank pages the way Google does. It assembles an answer and cites the sources that make that answer verifiable, so being citeable is a data-quality problem before it is a content problem.

At DevCommX we build signal-based GTM systems, and AI answer visibility is now part of that pipeline. The mechanics here are close cousins of what we cover in how to get cited by ChatGPT, but Claude is different in one important way: Anthropic has publicly documented exactly which bots it runs and what each one does. That documentation is the map. Most brands that are invisible in Claude are invisible for a boring, fixable reason, and this post walks through the specific reasons and the order to fix them in.

What Claude Actually Pulls: The Three Crawlers

In February 2026 Anthropic updated its support documentation to clearly separate its web crawlers by purpose. This matters because each bot answers a different question about your site, and blocking one does not block the others. The decisions are discrete. If you want to appear in Claude answers, you cannot treat "AI bots" as one undifferentiated group.

ClaudeBot is the training crawler. It collects web content that helps improve Anthropic's models over time. Content it reads may shape what Claude knows in general, but ClaudeBot is not what fetches a live citation when a user asks a question today. Claude-User is the user-initiated fetcher. When someone in Claude asks about your category and Claude decides to pull a page, Claude-User retrieves it in real time. Claude-SearchBot navigates the web to improve the quality of search-style results inside Claude, indexing and evaluating content so the assistant can return better sourced answers. There is also a fourth agent tied to Claude Code, but the three above are the ones that decide whether your brand shows up in a normal answer.

The user-agent strings are explicit. ClaudeBot identifies as "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com)" and Claude-SearchBot as "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Claude-SearchBot/1.0; +Claude-SearchBot@anthropic.com)". You can verify legitimate traffic against the source IP list Anthropic publishes at claude.com/crawling/bots.json. Log-level visibility into which of these bots reaches your pages, and which get a 200 versus a 403, is the single most useful diagnostic you can run.

The Three Crawlers Side by Side

Here is the practical breakdown of each bot, what it governs, and the mistake that quietly locks you out of it.

Fix the Technical Access Problem First

The most common reason a brand is absent from Claude is not weak content. It is that Claude-User or Claude-SearchBot receives a challenge page instead of your HTML. Cloudflare's bot-management and default WAF settings frequently challenge or block these agents silently, and the failure is invisible. There is no error in your dashboard. The bot simply gets a 403 or a JavaScript challenge, cannot read your content, and produces zero citations with no signal that anything is wrong.

Do three checks. First, open your robots.txt and confirm you are not disallowing ClaudeBot, Claude-User, or Claude-SearchBot, either directly or through a broad wildcard rule someone added to keep scrapers out. Anthropic honors standard Disallow directives and the Crawl-delay extension, so a single stray line removes you from consideration. Second, look in your CDN or WAF for bot rules that treat unknown user-agents as hostile, and add explicit allow rules for the Claude agents verified against the published IP list. Third, grep your server logs for the Claude user-agent strings and check the response codes. If you see 403s, that is your answer, and no amount of content work will fix it until the access issue is resolved.

One nuance worth stating plainly: allowing Claude-User and Claude-SearchBot while blocking ClaudeBot is a legitimate stance if you do not want your content used for training but do want to be citeable in live answers. Because the controls are independent, you can make that exact choice. Just make it deliberately rather than inheriting it from a copied robots.txt.

A robots.txt Starting Point You Can Copy

If you want to stay citeable in answers while opting out of training, the following pattern makes the intent explicit. It allows the two bots that fetch and index live pages and disallows only the training crawler. Adjust to your own policy, but never leave the decision to a vague wildcard.

Allow live citation, opt out of training: add User-agent: Claude-User with Allow: /, add User-agent: Claude-SearchBot with Allow: /, and add User-agent: ClaudeBot with Disallow: /. Allow everything: the simplest citeable posture is to add no Claude-specific Disallow rules at all and confirm no catch-all User-agent: * block is sweeping the bots up. Whichever you choose, remember that robots.txt is only half the story. A CDN rule can still block a bot that robots.txt technically permits, and the bot has to be able to reach robots.txt in the first place, so an aggressive IP-level block can stop it from ever reading your rules. Test the live response, do not just trust the file.

Why Third-Party Sources Decide Your Citations

Once Claude can read you, the harder truth is that your own website is not the main driver of how Claude describes your brand. Analyses of Claude citations consistently show that owned domains make up a small minority of what gets cited, with the large majority coming from independent editorial and third-party web content. In other words, what other credible sites say about you moves your citation rate faster than what you publish about yourself.

This reframes the work. Guest contributions on publications Claude already cites, inclusion in roundup and "best tools" articles, expert commentary in trade press, and accurate profiles on directories all feed the model a consistent picture. If your brand facts, your positioning, your category, your key numbers, are repeated the same way across your site, LinkedIn, Crunchbase, G2, and Wikidata, Claude speaks about you with confidence. When those sources disagree, the model hedges or drops the recommendation entirely. Consistency is not a branding nicety here. It is the input that determines whether Claude will commit to citing you.

This is also where an LLM-focused content program pays off. The structural work of making pages quotable and internally consistent is exactly what we lay out in the LLMO playbook for optimizing content for LLMs, and it applies cleanly to Claude.

How to Structure a Page Claude Will Cite

Claude rewards verifiable credibility over keyword density. The pages that get pulled share a recognizable shape, and you can engineer for it directly.

Lead with the answer. Put a direct, standalone answer in the first sentence of the page and the first sentence of every major section, then expand. Claude extracts the answer and cites the source, so a page that buries its conclusion under three paragraphs of preamble gives the model nothing clean to lift. Show your credibility signals. Named authors with real titles, primary data you actually gathered, dates that are visible on the page, and clear sourcing all raise the odds of citation. Reporting suggests brands with a rich set of structured attributes get cited several times more often than brands with almost none, so completeness compounds.

Write in questions. FAQ sections work unusually well because they mirror how people talk to Claude. Phrase the question the way a real user would type it, then answer it directly in the first line. Build topical depth. Claude trusts a site that covers a topic thoroughly more than one thin post. A single article about AI search will struggle against a competitor with a full cluster, so depth across a subject signals authority the model can lean on. Keep it fresh. Recent timestamps help in citation selection, so a visible publish or update date on every important page is worth the small effort.

Measure Whether Any of This Is Working

You cannot manage what you cannot see, and Claude visibility is measurable if you define the right numbers. Track three per prompt: mention rate, whether Claude names your brand in prose at all; citation rate, the percentage of prompts where Claude cites your actual URL as a source; and share of voice, how often you appear against a named set of competitors. Citation rate is the stronger signal, because being used as a source is a bigger trust marker than being mentioned in passing, and you should track it separately.

Build a fixed list of the prompts a real buyer in your category would ask, run them on a schedule, and record those three numbers over time. That gives you a before-and-after when you fix a robots.txt rule or land a guest post, which is the only way to know which lever actually moved. We go deeper on the tooling and cadence in our guide to measuring LLMO and AI visibility. The point is that Claude citation work is not a leap of faith. It is a loop you can instrument.

Why This Window Matters Now

Claude's reach makes this a category worth the effort, not a niche. Anthropic crossed roughly forty-seven billion dollars in annualized revenue in 2026, Claude sits in the range of hundreds of millions of monthly active users across web and mobile, and by early 2026 Anthropic had captured a majority of enterprise LLM spend, overtaking earlier leaders in that segment. The people asking Claude which vendor to shortlist are disproportionately the technical and enterprise buyers many B2B companies most want to reach.

Because Anthropic only recently documented the three-crawler model clearly, most competitors have not audited their access or restructured their pages for it. That gap is the opportunity. The brands that fix technical access, align their third-party footprint, and lead every page with an extractable answer will be the ones Claude reaches for while everyone else is still guessing why they never appear.

There is also a compounding effect worth naming. Once a page earns a citation in Claude, it tends to keep earning them for related prompts, because the model has already confirmed it as a source it can lean on. Early movers do not just win today's answer. They build a position that is harder for latecomers to dislodge, the same way an early backlink profile once compounded in traditional search. The cost of starting now is low. The cost of waiting is that a competitor becomes the default source Claude reaches for in your category, and you spend the following year trying to displace them.

Get Your Claude Visibility Audited

You can find out where you stand in about a minute. DevCommX runs a free AI Visibility Checker that shows how ChatGPT, Claude, and other assistants currently see your brand, whether your pages are even reachable by their crawlers, and which specific gaps are costing you citations. It is the same signal-based approach we use to take clients from setup to 40+ qualified demos in roughly six weeks, applied to AI answer visibility. Run the free checker, then book a GTM strategy call to turn the findings into a plan your team owns.

Further Reading

• Anthropic: Does Anthropic crawl data from the web, and how can site owners block the crawler?
• Anthropic: Published crawler source IP list (bots.json)
• Search Engine Land: Anthropic clarifies how Claude bots crawl sites and how to block them

FAQ

What are the three Claude crawlers and what does each do?

Anthropic runs three primary web crawlers. ClaudeBot collects content to help train and improve Claude's models. Claude-User fetches a live page in real time when a Claude user's question needs it. Claude-SearchBot indexes and evaluates content to improve search-quality answers inside Claude. Blocking one does not block the others, because each control is independent.

Why does my brand never appear in Claude answers?

The most common cause is technical, not editorial. Cloudflare or default WAF bot rules often challenge or block Claude-User and Claude-SearchBot silently, returning a 403 with no visible error, so Claude never reads your content. Check your robots.txt for stray Disallow rules and grep your server logs for the Claude user-agent strings and their response codes first.

Does my own website or third-party sites matter more for Claude citations?

Third-party sources matter more. Analyses of Claude citations consistently show owned domains are a small minority of what gets cited, with most coming from independent editorial content. Guest posts, roundup inclusions, and consistent profiles on G2, Crunchbase, and Wikidata move your citation rate faster than volume published only on your own domain.

How should I structure a page so Claude will cite it?

Lead with a direct, standalone answer in the first sentence of the page and each section, then expand. Add named authors, primary data, visible dates, and clear sourcing. Use FAQ formatting phrased the way real users ask Claude, and build topical depth across a subject rather than one thin post. Claude rewards verifiable credibility over keyword density.

Can I let Claude cite me without letting it train on my content?

Yes. Because Anthropic's crawler controls are independent, you can allow Claude-User and Claude-SearchBot so your pages remain citeable in live answers while disallowing ClaudeBot to opt out of training use. Set this deliberately in robots.txt using each bot's specific user-agent rather than inheriting a broad rule from a copied template.

How do I measure my visibility inside Claude?

Track three numbers per prompt: mention rate, whether Claude names your brand; citation rate, the percentage of prompts where Claude cites your URL as a source; and share of voice against named competitors. Run a fixed list of buyer-style prompts on a schedule and record the trend so you can tell which fixes actually moved your citations.