Blog

How to Rank in ChatGPT, Claude and Google AI Overviews: One Guide, Three Engines

← All posts
TL;DR
  • None of the three engines shows a ranked list. Each picks a few sources per answer, so the goal is to be one of them.
  • Shared work: let the crawlers in, get indexed where each engine looks, answer one question per page, and test with a fixed prompt set.
  • Different work: the bot to allow, the index behind it, and how much the company tells you. Google tells you a lot. OpenAI and Anthropic tell you little.

To rank in ChatGPT, Claude and Google AI Overviews, let each engine's crawler in, get into the index each one searches, and write pages that answer one question plainly. The steps are mostly shared. The bots and indexes are not.

All facts below carry the date they were checked: 31 Oct 2026.

What does "ranking" mean inside an AI answer?

There is no position one. Each engine writes an answer and attaches a few sources, or none. You can be cited (linked), mentioned (named in the text) or absent.

Two official statements set the tone. Google says a page must be indexed and eligible to show with a snippet to appear as a supporting link, and that meeting the rules doesn't guarantee crawling, indexing or serving (Google Search Central, checked 31 Oct 2026). OpenAI's site-owner guidance says ChatGPT orders search results using several factors and that placement is not guaranteed (OpenAI help center, as quoted by eSEOspace, checked 31 Oct 2026).

So split the job in two. Eligibility is yours to fix. Selection you can only influence.

What is the same for all three engines?

Let the right bot in. Every engine has a crawler or fetcher for search. A blanket Disallow, a CDN rule or a firewall setting can block it without anyone noticing. Google's own guide lists CDN and hosting rules next to robots.txt.

Be in the index the engine searches. Each engine reads from a search index first and writes second. If your page isn't in that index, no wording fixes it.

Put one clear answer under each heading. Lead each section with the answer, then add detail. Google says there is no need to cut content into tiny pieces, so this is for readers first. Nobody outside these companies has shown that it moves citations.

Say something other pages can't. Google calls unique, non-commodity content the factor most likely to matter over time. Its example: a first-hand account beats a generic tips list. OpenAI and Anthropic publish no equivalent advice, but they also choose among similar pages, so this is a reasonable bet for them too.

Measure with the same prompts every time. None of the three gives you a ranking report. More on that below.

How do the three engines differ?

ChatGPTClaudeGoogle AI Overviews
Bot to allow for searchOAI-SearchBotClaude-SearchBot (plus Claude-User for live fetches)Googlebot
Training-only botGPTBotClaudeBotGoogle-Extended (other Google AI systems, not Search)
Where results come fromOpenAI's own index plus outside providers; exact mix not publishedBrave Search, widely reportedGoogle's Search index
Official guidance on getting chosenCrawler docs and one short line on rankingCrawler help page onlyDetailed guide, updated 10 Jul 2026
Special optimisation neededNone publishedNone publishedGoogle says none
Reporting for site ownersNoneNoneSearch Console generative AI report

Details checked 31 Oct 2026.

How do you get cited in ChatGPT?

  1. Allow OAI-SearchBot. OpenAI says sites that opt out of this bot aren't shown in ChatGPT search answers, though they can still appear as plain navigational links. Changes can take about 24 hours to apply (OpenAI crawler docs, checked 31 Oct 2026). A missing rule counts as allowed, so look for a blanket Disallow: / that catches every bot. If you want to be explicit: 
    User-agent: OAI-SearchBot
Allow: /
  2. Keep GPTBot out of the decision. It collects training data. OpenAI documents each of its bots as a separate switch, so you can block training and stay in search.
  3. Check your infrastructure. OpenAI publishes IP ranges for its bots. Confirm your host or CDN isn't dropping them. ChatGPT-User fetches a page when a person asks about it, and OpenAI says robots.txt rules may not apply to those requests.
  4. Treat Bing as one input, not the whole story. Older guides describe ChatGPT search as a Bing front end. Peec, an AI-visibility vendor, reported in Sep 2026 that ChatGPT draws on OpenAI's own index plus several outside providers, including Google-sourced results and Microsoft's grounding platform (Peec, Search Engine Watch, checked 31 Oct 2026). OpenAI hasn't published the mix. A sitemap in Bing Webmaster Tools is still cheap, but it may not be the only lever.
  5. Look at who gets cited for your questions. Ask ChatGPT the questions your buyers ask and open the Sources panel. If the same list posts and review pages appear, those are the pages you need to be on.

How do you get cited in Claude?

  1. Allow Claude-SearchBot and Claude-User. ClaudeBot is for training. Anthropic says disabling Claude-SearchBot can reduce your visibility in search results, and disabling Claude-User can reduce it for user-directed lookups (Anthropic support, checked 31 Oct 2026). Use robots.txt only: Anthropic says blocking by IP can stop its bots from reading your robots.txt.
  2. Get into Brave. Claude's web search is widely reported to run on Brave Search. Anthropic's subprocessor list has included Brave for web search since March 2025, according to Xponent21. Third-party studies report a high overlap between Claude's cited URLs and Brave's top results, but the figures vary by study and year. Brave has no robots.txt token of its own and, per Scrunch (Mar 2025), follows the rules set for Googlebot.
  3. Expect Claude to skip search often. Claude doesn't search the web on every prompt. When it doesn't, the answer comes from training knowledge, which you can't edit. Keep your brand name and what you do consistent across the web.

How do you get into Google AI Overviews?

  1. Check eligibility first. The page must be indexed and eligible for a snippet. Test with URL Inspection in Search Console. nosnippet and max-snippet settings limit what Google can show, AI features included.
  2. Cover the sub-questions. Google says AI Overviews and AI Mode may run several related searches, which it calls query fan-out. A page that handles a topic well can surface for the sub-questions. Google also warns that mass-producing pages for each variant to manipulate results breaks its scaled content abuse policy.
  3. Put effort into non-commodity content. First-hand experience, your own data, a clear point of view.
  4. Skip what Google says to skip. Its guide to generative AI search (last updated 10 Jul 2026, checked 31 Oct 2026) says llms.txt files, content "chunking", rewriting for AI and special schema aren't needed for Google. It adds that seeking inauthentic mentions isn't as helpful as it seems.
  5. Watch the Search Console generative AI performance report. It's the only first-party report of the three engines.

How do you know if any of this worked?

Write 20 real questions your buyers ask. Run them in each engine on a fixed schedule, with the same settings, and log each as cited, mentioned or absent. Run each prompt twice: answers vary from run to run, so ignore one-prompt swings. Add the Search Console report for Google, and check your analytics source report for visits from each engine's domain.

What should you skip?

  • IP-blocking or IP-allowlisting Anthropic's bots instead of using robots.txt.
  • Adding llms.txt and expecting a Google effect. Google says it ignores the file.
  • One page per fan-out question.
  • Buying or faking mentions.
  • Treating any single percentage as law. Most citation studies come from tool vendors with something to sell.

What can't anyone outside these companies see?

OpenAI and Anthropic don't explain how sources are chosen. Google explains more but gives no weights. Everything beyond the crawler docs is observation, and some of it comes from people selling visibility tools. Re-check this guide when the engines change. ChatGPT's source mix and Claude's backends have both moved within the last year.

What should you do in week one?

  1. Open robots.txt and your CDN rules. Remove any rule that blocks OAI-SearchBot, Claude-SearchBot, Claude-User or Googlebot.
  2. Verify your site in Search Console and Bing Webmaster Tools, and submit your sitemap to both.
  3. Search your key page titles in Brave to see whether the pages are indexed.
  4. Rewrite the first lines of your top three pages so each answers its question directly.
  5. Run your 20-prompt baseline in all three engines and save the results.

For a ready-made sheet, see the free AEO audit template.

FAQ

How do I rank in ChatGPT?

Allow OAI-SearchBot, keep your pages indexed (including in Bing, as one input), and write pages that answer one question clearly. OpenAI says placement is never guaranteed.

Does blocking GPTBot remove me from ChatGPT search?

No, according to OpenAI's crawler documentation. GPTBot is the training crawler and OAI-SearchBot controls search. They are separate switches.

Does my Google rank decide whether Claude cites me?

Not directly. Claude's search is widely reported to run on Brave, so Brave visibility looks like the closer signal. Brave follows Googlebot's crawl rules, so a page Googlebot can't crawl is unlikely to reach Brave either.

Do I need llms.txt?

Not for Google. Google says it ignores the file and that it neither helps nor hurts. OpenAI and Anthropic haven't published support for it, so adding it is optional.

Sources (checked 31 Oct 2026)