Which Search Index Each AI Assistant Reads, and How to Get Into It

Here is the thing most AI SEO advice gets wrong: there is no single “AI index” to get into. Each assistant answers from a different underlying search index, and the way you get into one has almost nothing to do with the next. ChatGPT leans on Bing. Claude reads Brave. Perplexity runs its own. Gemini and Google AI Overviews use Google. So the question is not how to get indexed by AI; it is how to get indexed by each engine, because the levers are different and some of them surprise people. This is the verified map and the exact move per engine.

The index behind each assistant

Read that table top to bottom and the strategy falls out of it. Two engines share Bing, one rides Google, one is on Brave, and one built its own. There is no shortcut that covers all five, which is also why a brand can be cited everywhere on one assistant and invisible on another. We saw the same split in the data when we compared citations across five engines and found they agree on only a small fraction of their sources.

ChatGPT and Copilot: the Bing family, where IndexNow wins

ChatGPT search is built on a combination of OpenAI’s own crawl and the Bing index it has long drawn on, so you have two jobs. First, let OAI-SearchBot reach you, because OpenAI’s crawler is what surfaces your pages inside ChatGPT search; block it and you are out even if you rank in Bing. Second, get into Bing fast, because that index is the foundation underneath the answer.

The fast way into Bing is IndexNow. It is a protocol that lets you ping a search engine the moment you publish or update a page, instead of waiting for a crawl. One submission is shared across every participating engine: Bing, Yandex, Seznam, Naver, and Yep. Set it up once, usually through Bing Webmaster Tools or a plugin on platforms like WordPress and Cloudflare that ship native support, and your fresh URLs reach the Bing index in minutes rather than days. Because ChatGPT and Copilot both read Bing, IndexNow is the single highest-leverage submission move you can make for either one. The full per-engine playbook for ChatGPT, measuring where you stand and improving it, is in our guide to tracking and ranking on ChatGPT.

Claude: the one that breaks your playbook

Claude is the engine people get wrong most often, because none of the above applies to it. Claude’s web search retrieves from Brave Search, which Anthropic lists as a subprocessor and which runs its own independent index. Your Bing IndexNow setup does nothing for Claude. Your Google Search Console does nothing for Claude. If you only optimize for the big two, you can be completely absent from Claude and never know why.

What actually moves Brave is unusual, and it comes straight from Brave’s own crawler documentation. Brave does not run a webmaster tool, so there is no add-your-site form. Its crawler does not even advertise a distinct user agent, so you cannot single it out in robots.txt. And it will not crawl a page that Googlebot cannot: if your robots.txt blocks Googlebot, Brave skips you too. So the path into Claude is to keep Googlebot access clean and earn a genuine Brave ranking, helped along by Brave’s opt-in Web Discovery Project. The only submission Brave offers is a re-fetch request, which exists to apply a delisting or clear a dead page, not to add a new site.

Perplexity: its own index, no shortcut

Perplexity does not borrow anyone’s index; it builds its own with PerplexityBot. There is no submission tool and no IndexNow path, so the levers are the plain ones: let PerplexityBot crawl you, and keep the pages you want cited current, because Perplexity weights freshness heavily and has no fixed knowledge cutoff. A maintained, recently dated page is the closest thing to a submission Perplexity offers, and we go deeper on earning those citations in our guide to getting cited on Perplexity.

Google: Search Console, not IndexNow

Gemini, AI Overviews, and AI Mode all ground on the main Google Search index, so the path in is ordinary Google SEO: be crawlable by Googlebot, submit a sitemap, and use Google Search Console to request indexing. One thing to know: Google does not participate in IndexNow, so the ping that accelerates Bing does nothing for Google. Google still wants sitemaps and Search Console. There is also no separate AI token to manage here, because AI Overviews draw on the same index as classic Search. Indexing is only the entry fee: Gemini, AI Overviews, and AI Mode are three separate surfaces that cite different sources, which we break down in our guide to getting cited in Google AI Overviews, AI Mode, and Gemini.

The reality check: indexed is necessary, not sufficient

Getting into these indexes is the price of entry, but do not expect the click flood that classic search used to bring. AI crawlers take a great deal and send very little back. Cloudflare Radar’s AI Insights tracks a crawl-to-referral ratio, how many pages a platform fetches for each visitor it sends you, and the numbers are stark: AI platforms crawl hundreds to thousands of pages per referral, with Anthropic the most lopsided, while Google search still returns a visitor for roughly every few crawls. Cloudflare’s own write-up frames the trend plainly: more crawls, fewer referrals.

The practical reading is not to block the crawlers; it is to measure the right outcome. Being indexed makes you eligible to be cited, and a citation is the win on these platforms, not the trickle of referral traffic behind it. Track whether each engine cites you for the prompts that matter, per engine, rather than waiting for an AI traffic spike that the crawl-to-referral math says is not coming.

The per-engine checklist

• ChatGPT: allow OAI-SearchBot, set up IndexNow and Bing Webmaster Tools.
• Copilot: Bing Webmaster Tools and IndexNow.
• Claude: keep Googlebot access clean, earn Brave ranking; there is nothing to submit.
• Perplexity: allow PerplexityBot, keep key pages fresh and dated.
• Gemini and AI Overviews: Googlebot access, a sitemap, and Search Console.

The common thread is reachability: every one of these starts with the right crawler being able to fetch your pages, which is the single most common silent failure in AI visibility. We keep a full reference of every AI crawler and what blocking it costs, and our robots.txt checker tells you what each bot can reach right now; if you are starting from scratch, crawl access is the first thing to verify. Once you are reachable and submitted where submission exists, the rest is the same earned-citation work, engine by engine, that we lay out in our AI search playbook.

Last reviewed June 2026. The index relationships behind these assistants change as the platforms strike and drop partnerships; the vendor documentation linked above is the source of truth, and we re-check this map each quarter.