Getting Cited by Claude: Why Brave Search Controls Your Visibility, and How to Fix It

Your site ranks well on Google. Bing knows you exist. And yet, when someone asks Claude a question in your field with web search on, it cites everyone but you. This is not a bug, and it is not necessarily a content-quality problem: Claude simply is not looking where you are.

I hit this exact situation with this very site, in July 2026. The cause fit in one line: my site was missing from the Brave Search index, the engine Claude leans on to search the web. I submitted one evening, and I was cited as the first source the next day. That test, and the whole mechanism behind it, is what I break down here.

The short version (for the impatient)

• Claude's web search runs on the Brave Search index, not Google and not Bing.
• According to a Profound analysis (2025), 86.7% of the sources Claude cites match Brave's top organic results.
• A site absent from Brave's index is invisible to Claude, even sitting at #1 on Google.
• Two-minute check: type site:yourdomain.com into search.brave.com.
• The most reliable way in: rank on Google and Bing. Brave harvests your URL the moment a Brave user searches there, and that is what feeds Claude.
• Your levers: manual URL submission to Brave, a robots.txt open to Claude-SearchBot and Claude-User, then genuinely citable content (structure, sourced stats, freshness).

Why Claude ignores your site, even when you rank #1 on Google

The short answer: your Google position never enters Claude's search loop. When Claude needs fresh information, it queries an external engine, and that engine is neither Google nor Bing.

A useful reminder: a model like Claude only "knows" what it saw during training. For everything else, the system goes and retrieves live sources from an index, then writes its answer from them, citing them. I broke down that retrieval mechanism in my guide on how LLMs actually work for SEO.

The consequence is brutally simple: every AI assistant has its own pool of sources. Being absent from the pool means not existing for that system, whatever the rest of your SEO looks like. For Claude, that pool is the Brave index.

🎯 What this changes for you. Before you talk about "optimizing for AI," ask the prerequisite question: am I even in the index this system draws from? Everything else comes after.

The evidence that Claude runs on Brave Search

Anthropic has never officially named its search provider. But the trail of evidence, documented since March 2025, leaves little room for doubt.

On March 20, 2025, Anthropic switched on web search in Claude. The very next day, developer Simon Willison noticed that "Brave Search" had just been added to Anthropic's list of subprocessors, and that the same query returned identical citations in Claude and in Brave. He even spotted a parameter named BraveSearchParams inside the search function. TechCrunch confirmed the analysis the same day.

Since then, the correlation has been measured. A Profound analysis (2025) compared the sources Claude cites to Brave's top organic results: 86.7% overlap (13 out of 15), with a p-value below 0.0001. In other words, this is not chance. More striking still: where ChatGPT heavily reorders Bing's results (only 26.7% alignment in the same analysis), Claude hands back Brave's rankings almost untouched.

The pattern was just corroborated at a larger scale. A July 2026 analysis by MERJ, across roughly 35,000 Claude citations on 400 queries, found that 79.2% of the URLs Claude cites sit in Brave's top 10 for the matching query, against just 34% for Google's top 10. Same conclusion as Profound, far bigger sample, and a blunt contrast: Brave predicts Claude's citations far better than Google does.

And that pool is anything but marginal. Brave claims an independent index of 40 billion pages, refreshed at more than 100 million pages a day (February 2026), and says it powers real-time search for most of the ten largest LLMs on the market.

🎯 What this changes for you. Your visibility in Brave Search almost mechanically determines your eligibility to be cited in Claude. It is one of the rare AI systems where the lever is this direct and this predictable.

My test: invisible one night, cited the next day

Here is the test that made me write this article. In July 2026, I asked Claude, with web search on, what it knew about "Raouf Douihech, SEO consultant." It cited several sources: LinkedIn, a few directories. But not my site, even though it was indexed on Google and Bing and ranking for my own name.

The next reflex explained everything: site:raoufdouihech.com on search.brave.com, zero results. My site did not exist in Brave's index. So it did not exist for Claude. My Google ranking could do nothing about it.

I submitted the homepage through Brave's form that evening. The next day, the page was indexed, and asking Claude the exact same question again, my site came back as the first cited source. No other variable moved: no content edited, no backlink earned, no tag added. The only difference between "invisible" and "first source" was presence in Brave's index.

I told the story of this test in a LinkedIn post that got plenty of SEOs talking: most were discovering Brave's role for the first time. That blind spot is exactly what this article sets out to close.

🎯 What this changes for you. A one-minute test (site: on Brave) can explain months of invisibility on Claude. Do it before any audit, any tool purchase, any "AI-special" redesign.

Check in 2 minutes whether Claude can cite you

The diagnosis comes down to three moves, in this order:

1. Test your Brave indexing: type site:yourdomain.com into search.brave.com. Look at the number of pages, not just whether the homepage shows up: your strategic pages (services, guides) need to be there.
2. Ask Claude your customers' questions, with web search on: "best [your service] in [your city]," "how to choose [your product]." Note who gets cited, and whether you are in there.
3. Cross-check: if the competitors Claude cites are the ones ranking for the same queries in Brave Search, you have your confirmation of the mechanism, and your action plan.

Zero results on the site: query? You are invisible to Claude, and the rest of this article is your roadmap.

How to get indexed on Brave Search

First thing to know: Brave has no equivalent of Search Console. No sitemap submission, no coverage report, no IndexNow. So how do pages actually get into the index? A July 2026 technical breakdown by MERJ, built from reading Brave's open-source Web Discovery Project code, maps three distinct doors:

• The query channel, your cheapest lever: when a Brave user runs a search on Google, Bing, Yahoo or DuckDuckGo where you already rank, Brave harvests your URL straight off that results page. A single searcher is enough. In MERJ's words, ranking on the incumbents is "the cheapest and most controllable" way in, which is why your classic Google and Bing SEO quietly underpins your Brave visibility, and downstream, your Claude citations.
• The page channel: roughly 20 opted-in Brave browser users, on different networks, visiting your page. A strong quality signal, but a high bar to hit on purpose.
• Brave's own crawler: server-side crawling that, per MERJ, only fetches what Googlebot can fetch. If your page is not crawlable by Googlebot, Brave's bot will not touch it either.

So where does the famous submission form fit? Here is the nuance worth knowing: reading the code, MERJ concludes Brave's URL submission form triggers a re-fetch, not first-time discovery. My own test looks like it says otherwise (I submitted a page that was absent from the index, and it showed up within 24 hours), but the two reconcile cleanly: my site already ranked on Google and Bing, so the query channel had almost certainly primed the ground, and the submission pushed the fetch through.

The practical playbook is the same either way:

• Earn classic visibility first: rank on Google and Bing for the queries your prospects actually type. That is what seeds Brave on its own.
• Submit your key URLs through Brave's URL submission form, homepage first, then the pages that answer what people ask Claude. In my case, the homepage appeared in less than 24 hours.
• Re-check with site: a few days later, and resubmit the pages that matter but are still missing.

🎯 What this changes for you. Stop treating Brave as a separate SEO project. Your most reliable way in is the SEO you already do for Google and Bing: rank there, and a single Brave user's search carries you into the very index Claude reads.

Open your robots.txt to Anthropic's crawlers

Second door to check: your own barriers. Anthropic documents three distinct crawlers, each controllable separately in your robots.txt:

• ClaudeBot: collects public content to train future models. Blocking it does not affect your visibility in Claude's answers.
• Claude-SearchBot: crawls the web to improve the quality and relevance of Claude's search results. Do not block it if you want to be cited.
• Claude-User: fetches a page live when a user asks Claude to read it. Blocking it stops Claude from reading your pages on demand.

The pragmatic stance for visibility: let Claude-SearchBot and Claude-User through. The ClaudeBot case (training) is a separate editorial choice, with no impact on citations. All three respect robots.txt, and blocking one does not block the others.

An increasingly common trap: your CDN or your firewall. Anti-AI-bot protections (Cloudflare and the like) are spreading and can block these crawlers before your robots.txt even gets a say. Check your server logs: if Claude-SearchBot never shows up, look for the block upstream.

One subtlety specific to Brave: its discovery crawler is deliberately indistinct, with no dedicated user-agent, and (per the same MERJ analysis) it only fetches what Googlebot can fetch. Two takeaways: a blanket "block all unknown bots" rule quietly locks Brave out, and anything that must be found should be server-rendered, because Brave's discovery layer does not run JavaScript.

Being indexed isn't enough: becoming citable

Indexing gets you into the pool. You still have to be chosen. Two layers to work on.

First layer: rank on Brave for the queries Claude generates. Since Claude hands back Brave's rankings almost untouched (the 86.7% from Profound above), the fundamentals pay off directly: content that answers the question, clear titles, authority, internal linking. The healthy SEO you already do, applied to one more index.

Second layer: be the kind of content a model cites. And here, research gives numbers. The founding GEO paper (Princeton, published at KDD 2024) measured, across 10,000 queries, that citing your sources, adding statistics and adding expert quotes lifts visibility in generative answers by 30 to 40%. Concretely:

• Source every claim: a fact that is attributed, dated and linked is more "citable" than an opinion.
• Structure for the machine: a direct answer at the top of each section, faithful H2/H3, tables, lists.
• Date and refresh: visibly up-to-date content reassures a system that favors reliability.
• Get corroborated: Claude cross-checks sources. A consistent LinkedIn profile, third-party mentions, serious directories all reinforce your entity.
• Cover the whole topic: AI systems split a question into sub-questions. A page that only answers the head keyword misses most of the entry points.
• Stay balanced: content that acknowledges its limits is structurally more quoted than a one-sided pitch.

💼 Want to know where you stand? I audit your visibility on Claude, ChatGPT and AI Overviews (which pool indexes you, sub-question coverage, citability), with a prioritized action plan. Discover my SEO and GEO services.

Why Claude's audience is worth the effort

In raw consumer volume, Claude weighs less than ChatGPT. But looking at raw volume would miss the point: Claude's audience is overwhelmingly professional.

According to Menlo Ventures' State of Generative AI in the Enterprise report (late 2025), Anthropic captures 40% of enterprise LLM spend, ahead of OpenAI (27%), when it held only 12% in 2023. In the code generation market, its share reaches 54%. All of this on an enterprise LLM market that grew to 8.4 billion dollars.

Translation for you: Claude's users are developers, consultants, decision-makers. If you sell B2B, being cited by Claude puts you in front of people who buy, not in front of curiosity traffic. And that traffic is measurable: in GA4, look for the claude.ai referral in your acquisition reports.

The map of indexes: every AI has its own pool

Brave for Claude is just one case of a general rule: "optimizing for AI" in the singular does not exist. Each system draws from its own pool, and the entry prerequisite changes every time.

System	Source pool	The entry prerequisite
Claude	Brave Search index	Be indexed on Brave (site: + manual submission)
ChatGPT (Search)	Bing index	Be indexed on Bing (Bing Webmaster Tools, IndexNow)
Google AI Overviews / AI Mode	Google index	Healthy Google SEO, sub-question coverage
Perplexity	Proprietary index + on-demand crawl	Allow PerplexityBot, freshness, authority

One honesty note, because you will read all sorts of things: this work changes nothing about your Google ranking. GEO serves your visibility in AI engines, classic SEO serves Google: two mechanics, one shared foundation of quality. For the big-picture view of that mechanism, see my guide on how LLMs work applied to SEO.

FAQ

Conclusion

Getting cited by Claude comes down to a short chain: be in the Brave index, let Anthropic's crawlers through, and publish citable content. The first link is checked in one minute and fixed in one evening. A lever this direct is rare in SEO.

Three reflexes to keep:

1. Check the pool: site:yourdomain.com on search.brave.com, today.
2. Open the doors: submit your key pages to Brave, check robots.txt, CDN and firewall.
3. Become citable: direct answers, sourced and dated facts, the topic covered end to end.

And if you want someone to run that diagnosis for you, platform by platform, with a prioritized action plan: here's how I work on SEO and GEO visibility.