Does Claude use Google Search?

No. Anthropic added Brave Search to its subprocessor list on March 19, 2025, the day before announcing Claude web search, and independent testing found Claude's citations overlapping Brave's top results roughly 87% of the time. Anthropic has not confirmed the arrangement publicly, so treat it as strongly evidenced rather than official — but it is clearly not Google.

Can I submit a sitemap to Brave Search?

No. Brave has no Search Console equivalent — no sitemap upload, no coverage reports, no crawl error exports. Pages are submitted individually at search.brave.com/submit-url, which is also the route for asking Brave to re-fetch a page that has changed. Because it is one URL at a time, prioritize your homepage, your bio page, and the neighborhood pages for areas you actually work.

Why do I rank on Google but never get mentioned by AI assistants?

Usually because they are not reading Google. ChatGPT runs its own search crawler, Claude's search appears to run on Brave, and Perplexity maintains its own index — so strong Google rankings do not automatically place you in any of them. It is a discovery problem rather than a ranking problem, and the fix is making sure each index has actually crawled you.

Should I block AI bots in robots.txt?

Not indiscriminately, because the bots do different jobs. OpenAI documents GPTBot as the crawler for model training, while OAI-SearchBot is what surfaces you in ChatGPT's search results — blocking the second removes you from ChatGPT entirely. If your goal is staying out of training data, block the training crawler and leave the search crawlers alone.

How do I make sure Brave can crawl my site?

Make sure Googlebot can. Brave's own crawler documentation states that if a page is not crawlable by Googlebot, Brave's bot will not crawl it either, and that Brave deliberately does not advertise a differentiated user agent. So the Googlebot rules in your robots.txt are the lever — a Brave-specific rule is not.

Does submitting a URL guarantee it gets indexed or cited?

No. Submission places the URL in a crawl queue and nothing more. The page still has to be worth indexing, and being indexed is separate again from being cited in an answer. Discovery is a prerequisite rather than a strategy — what earns the mention is consistent facts and corroboration across independent sources.

Does Claude Use Google? What Each AI Assistant Actually Searches

The short answer, and how it's known

Claude's web search does not run on Google.

The evidence is circumstantial but strong, and it comes from Anthropic's own legal disclosures rather than a press release. On March 19, 2025, Anthropic added Brave Search to the subprocessor list in its Trust Center — the list of third parties it shares data with to deliver a service. Claude web search was announced the following day — and the subprocessor entry linked to that very support article. Simon Willison spotted the entry and corroborated it two ways: Claude's web_search function carries a parameter named BraveSearchParams, and running the same queries through both produced heavily overlapping results. TechCrunch reported a measured overlap of roughly 87% between Claude's citations and Brave's top non-sponsored results.

The honest caveat: Anthropic has never confirmed a web-search provider on the record. This is a well-evidenced inference from a subprocessor list, a parameter name, and result comparison — not an announced partnership. It is also not static: a second web-search subprocessor, the search database TurboPuffer, was added in May 2026, so "Claude runs on Brave" is a simplification of a stack that is still moving.

What is not in doubt is the part that affects you: it isn't Google.

Nobody searches the same index

Agents are told to "optimize for AI search" as though it were one destination. It is four or five separate ones, each reading from a different place:

Google AI Overviews and AI Mode — Google's index. If you rank on Google, you are already in the pool here.
ChatGPT search — OpenAI runs its own search crawler. OpenAI's own documentation describes OAI-SearchBot as "used to surface websites in search results in ChatGPT's search features," separate from the crawler used for model training. Bing has also historically been in the mix, which is why IndexNow still matters.
Claude — Brave Search, per the above.
Perplexity — its own index. Perplexity documents PerplexityBot as "designed to surface and link websites in search results on Perplexity."
Microsoft Copilot — Bing.

So an agent who is genuinely strong on Google may be entirely absent from Claude, because Claude never asked Google. That is not a ranking failure — it is a discovery failure, and it has a completely different fix.

Brave is the odd one out, and it's the one with no dashboard

Here is where it gets practical, and slightly annoying.

Google gives you Search Console. Bing gives you Webmaster Tools. Both let you submit a sitemap once and see coverage reports afterward. Brave gives you a form.

There is no Brave equivalent of Search Console: no sitemap upload, no index coverage report, no crawl error export, no query performance view. What exists is a single-URL submission page at search.brave.com/submit-url, which Brave's own crawler documentation points to for getting a page fetched or re-fetched.

One URL at a time. For a site with ninety pages that is genuinely tedious, which is exactly why almost nobody does it — and why doing it is worth something. You are not competing against agents who did this carefully. You are competing against agents who have never heard of it.

The same form is the route for a page that has changed — a re-fetch request — which matters after a rebuild or a site migration, when Brave may be holding an old version of a page that no longer exists.

The rule that does most of the work: Googlebot governs Brave

Before you spend an evening pasting URLs, there is one line in Brave's crawler documentation that decides whether any of it will work:

"if a domain or page is not crawlable by Googlebot, then Brave Search's bot will not crawl it either."

Brave also says plainly that its crawler "does not advertise a differentiated user agent because we must avoid discrimination from websites that allow only Google to crawl them."

Two useful conclusions follow. First, adding a special rule for a Brave-specific bot token in robots.txt is not the lever people think it is — the Googlebot rules are the lever. Second, if anything in your setup blocks or restricts Googlebot, you have silently locked yourself out of Claude as well, and no amount of URL submission will fix it.

Brave notes one more thing worth internalizing, because it trips up people on every search engine: robots.txt does not remove a page from an index. Blocking a crawler stops the fetch; the noindex directive is what delists.

Check what you are blocking before you submit anything

Plenty of agents on hosted website platforms are blocking the crawlers that matter without knowing it — sometimes because a well-meaning "block AI bots" toggle got switched on.

Open yourdomain.com/robots.txt in a browser and read it. Then understand the distinction almost nobody draws, because it changes what you should actually block:

OAI-SearchBot — surfaces you in ChatGPT's search results. Blocking this removes you from ChatGPT. Almost certainly not what you want.
GPTBot — per OpenAI's docs, crawls "content that may be used in training our generative AI foundation models." Blocking this is a defensible choice, and it does not remove you from ChatGPT search.
ChatGPT-User — fetches a page live when a user's question sends it there. Blocking it means ChatGPT can't open your site mid-conversation.
PerplexityBot — Perplexity's search crawler; Perplexity states it "is not used to crawl content for AI foundation models."
Googlebot — governs Google, and by Brave's own policy, Claude too.

The training-versus-search distinction is the whole point. An agent who wants their content out of model training but still wants to be recommended by ChatGPT can have both. An agent who blanket-blocks "AI bots" has thrown away the recommendations along with the training, which is the opposite of what they were told they were doing.

What to submit to Brave, in order

Since it is one at a time, spend the effort where it pays. Submit in this order and stop when you run out of patience — the first five matter far more than the next fifty.

Your homepage. It is the page most likely to be fetched as the canonical statement of who you are.
Your about or bio page. This is the page carrying the facts that resolve you as a person — name, brokerage, market, credentials — which is why keeping those facts identical everywhere is worth doing before you submit, not after.
Your neighborhood pages. The ones targeting areas you actually work. These are the pages that answer "who's the agent for this area," so they are the ones you want in the index Claude reads.
Your best two or three guides. Whatever you have that genuinely answers a question a client asks.
Anything you just changed. Re-fetch after a rebuild, a rebrand, or a URL change.

Then do the equivalent everywhere else: submit your sitemap in Google Search Console, claim Bing Webmaster Tools (which imports from Search Console in a couple of clicks), and use IndexNow for the Bing-fed side. None of this is glamorous and all of it is a one-time afternoon.

What submitting a URL does not do

Being honest about the ceiling here, because the tactic gets oversold.

Submitting a URL puts it in a crawl queue. That is all it does. It does not guarantee the page gets indexed, it does not guarantee the page gets cited, and it certainly does not make an assistant recommend you. A thin, duplicated, or genuinely unhelpful page that gets crawled promptly is still a thin page.

Discovery is a prerequisite, not a strategy. Being in the index is what makes it possible for you to be named; what actually gets you named is corroboration across independent sources — consistent facts, real reviews, profiles that agree with each other, and content only someone in your market could have written.

The right way to hold it: this post fixes the plumbing. If the assistant has never seen your site, nothing else you do can work. Once it has, everything else starts to. The free playbook is the rest of it in order, and the AI visibility checker will tell you where you currently stand.