How to rank in Brave Search (and get cited by Claude)

Brave Search answers from its own index, which grows from a crawler that follows the rules your site sets for Googlebot and from anonymous signals sent by Brave browser users who opted into the Web Discovery Project. There is no console or sitemap submission. To rank, let the crawler in, earn real visits, and be the best answer. Claude's web search appears to draw on Brave, so the same work affects whether Claude can find and cite you.

The facts below were checked against the published pages of Brave and others on 12 October 2026.

How does Brave Search find pages?

Brave says its results come from its own index, built from scratch rather than borrowed from another engine (What is Brave Search?). Its API page puts that index at more than 40 billion pages, kept fresh by over 100 million page updates a day; those are Brave's own figures (Brave Search API). Pages reach it through two routes.

The crawler. Brave's help page for site owners (Brave Search crawler) makes four points worth knowing:

• The crawler does not identify itself with a separate user agent, because Brave wants to avoid being shut out by sites that allow only Google to crawl them.
• If Googlebot is not allowed to crawl a domain or page, Brave's crawler will not crawl it either.
• robots.txt is not used to keep a page out of Brave's index. To remove a page, add a robots noindex directive; Brave then has to fetch the page again to see it, and you can submit the page for a re-fetch.
• The crawler is partly guided by information from users, which is the second route.

The Web Discovery Project. This is an opt-in feature of the Brave browser that sends anonymous search and browsing data from users who turned it on. Brave says it helps build the independent index and keep results relevant (Brave Search and the Web Discovery Project). Its privacy rules shape what it can report: for a URL to be sent, it has to be visited independently by a large number of people, and the project discards suspicious queries, URLs of pages their creators marked as non-indexable, and pages that are not public or require a login.

The consequence for site owners is simple. Brave leans on what people actually visit to find pages, and it reads them with the permissions you give Google.

What you can control

1. Let the crawler in. Read your robots.txt as Googlebot would (the group for Googlebot if you have one, otherwise the group for *), because Brave applies whatever Google is allowed. A section you hide from Googlebot is hidden from Brave too. Then look at what sits in front of your site. Brave's crawler does not announce itself separately, and its help page does not describe a way to verify its requests, so a firewall or bot-protection rule that only admits crawlers it can verify as Google's may also shut out Brave. If your setup works that way, it is worth testing.

2. Use noindex, not robots.txt, to keep pages out. Brave does not use robots.txt to decide what is indexed. Pages you want out of Brave's results, such as internal search results or staging copies, need a noindex directive, and they need to be crawlable so the directive can be seen. This is also how Google works, so one rule serves both engines (Block indexing with noindex).

3. Keep important pages public. Pages behind a login, and pages you marked as non-indexable, are excluded from what the Web Discovery Project reports.

4. Earn real visits. Because a URL is only reported once many people have visited it independently, a page nobody reads is invisible to that route, however good it is. The work that brings readers anyway (ranking in other engines, a newsletter, communities where your audience spends time, links from other sites) is also what puts a page in front of Brave users who share data. Brave users are a fraction of any audience, so the more visits a page earns overall, the sooner enough of them come from people taking part.

5. Ask for a re-fetch after a change. Brave points site owners to its submission form to have a page fetched again, for example after adding noindex. Treat it as a way to get a known page re-read, not as a way into the index.

6. Be the best answer. Brave's index is only the entry ticket; ranking still depends on how well the page answers the search. Brave's AI answers summarise web results and always cite their sources, according to its search page, so a page with a clear, specific answer near the top is easier to quote as well as to rank.

How to check whether Brave has your page

There is no Brave report to look at, so check by hand. Search search.brave.com for the page's exact title in quotes, or for a distinctive sentence from it. If a page that ranks in Google is missing from Brave after several weeks, go back through points 1 to 4: crawler access first, then whether the page has had enough real visitors to be reported.

What does Brave have to do with Claude?

Claude's web search appears to draw on Brave Search, although Anthropic has not published a full description of how it searches. The evidence, as of 12 October 2026:

• In March 2025, Anthropic's list of subprocessors added Brave Search, linking to its help page for Claude's web search, as documented by Simon Willison at the time. In the same write-up, Claude's search tool described its parameters as "BraveSearchParams", and the ten results Claude cited for a test search matched Brave's results for the same query exactly.
• Brave's own API page calls it the leading search tool for applications that use Claude through MCP.
• We could not load Anthropic's current subprocessor list on that date, because the page needs a full browser, so we have not confirmed that the listing is unchanged.

Two cautions follow. Being in Brave's index does not mean Claude will cite you: Claude chooses what to quote, and it may use other sources too. And everything above about crawling and visits applies whether or not a given answer comes from Brave. The practical takeaway is that a site invisible to Brave is likely to be invisible to Claude's web search as well, so the checks in this playbook are worth the half hour.

What AutoPSEO can and cannot do here

AutoPSEO does not measure Brave rankings. Its search reports come from Google Search Console, GA4 and Bing Webmaster Tools, and IndexNow, which AutoPSEO uses to notify Bing and other engines of new pages, does not include Brave: the engines listed on IndexNow's own site are Amazon, Bing, Naver, Seznam.cz, Yandex and Yep (IndexNow FAQ).

What it can show is the result at the other end:

• The LLM Traffic report groups GA4 sessions referred by AI assistants, Claude among them (visits from claude.ai), so you can see whether citations turn into visitors.
• Prompt Explorer, one of the research tools, shows what ChatGPT, Claude, Gemini and Perplexity answer to a prompt and which sources they cite. It runs on research credits.
• The technical indexing playbook covers the robots.txt, noindex and canonical checks that matter for every engine, Brave included.

The content engine helps with point 6. Each article it writes answers one search, opens with a direct answer of no more than 60 words by default, and is delivered to your own site as an ordinary page, so it is open to every crawler your site lets in (content engine).