How to rank in ChatGPT, Perplexity, Gemini and Claude answers

How do you rank in ChatGPT and other AI answers?

You rank in ChatGPT and other AI answers by being reachable by each engine's search crawler and by publishing the page that answers the question most clearly; nothing else is documented by the engines themselves. When an assistant searches the web, it pulls a handful of pages from its own index or a partner's, writes an answer, and cites the pages it used. Your job is to be in that index and to be the most useful page in it.

There is no "position 1" in a chat answer the way there is on a results page. "Ranking" here means one of two things: the answer names your brand or product, or it cites your page as a source. Both can happen when the engine searches the web. Only naming can happen when it answers from memory, which depends on what the model learned in training and changes slowly.

This guide is the engine-by-engine version. If you only care about Google, read how to get cited in AI Overviews. For the robots.txt lines themselves, see how to allow or block AI crawlers. For the wider method, see our generative engine optimization guide.

Which crawler or index does each AI engine use?

ChatGPT search uses OpenAI's OAI-SearchBot, Perplexity uses PerplexityBot, Claude uses Claude-SearchBot, Gemini grounds on the Google Search index, and Microsoft reports Copilot citations through Bing. Each company also runs a separate bot for model training and one for fetching a page when a person asks, and those do not decide whether you appear in search answers.

A practical warning: robots.txt is only half of access. If your CDN, firewall or bot-protection service challenges these crawlers, they cannot read the page no matter what robots.txt says. OpenAI and Perplexity both publish IP lists so you can allow their bots at the firewall.

How to show up in ChatGPT answers

To show up in ChatGPT search answers, allow OAI-SearchBot in robots.txt and in your CDN, then publish pages that answer the questions your buyers ask. OpenAI's crawler overview says OAI-SearchBot is used to surface websites in ChatGPT's search features, and that sites which opt out will not be shown in ChatGPT search answers.

• OAI-SearchBot decides search eligibility. OpenAI says a robots.txt change takes effect in about 24 hours, and publishes its IP ranges at openai.com/searchbot.json.
• GPTBot collects content that may be used to train OpenAI's foundation models. OpenAI documents it as independent of search: you can allow OAI-SearchBot and disallow GPTBot.
• ChatGPT-User visits a page when a user's request calls for it. OpenAI says it is not used to decide whether content may appear in search, and that robots.txt rules may not apply to it because the visit is user-initiated.

When ChatGPT does not search, it answers from what the model learned in training. You cannot edit that directly. What you can influence is the web it learns from and searches later: consistent facts about your product on your own site and on the pages that already rank for your category.

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

How to rank in Perplexity

Perplexity searches the web for every answer and shows numbered citations, so the way in is to allow PerplexityBot and be among the best pages on the open web for the question. Its help center describes three steps: understand the question, search the internet in real time, then summarise the most relevant findings with numbered citations linking to the original sources.

Perplexity's crawler documentation lists two agents. PerplexityBot is designed to surface and link websites in Perplexity's search results, and Perplexity recommends allowing it. Perplexity-User fetches a page when a person asks a question and, because the fetch is user-requested, generally ignores robots.txt. Perplexity says neither is used to collect content for training foundation models, so there is no training trade-off to weigh here.

Because Perplexity searches every time, it is the engine where better pages show up fastest. If you are missing, look at the domains it cites for your question: those are the pages you are competing with, and often the review sites and comparison lists where your product should be listed accurately.

How to show up in Gemini answers

Gemini Apps ground their answers on content from the Google Search index, so you need to be indexed in Google and not blocked through the Google-Extended token. Google's common crawlers page says Google-Extended controls whether crawled content may be used to train future Gemini models and for grounding, which it defines as providing content from the Google Search index to the model at prompt time, in Gemini Apps and Grounding with Google Search on Vertex AI.

Two details matter. First, Google-Extended does not affect your inclusion or ranking in Google Search, and Google's AI features page says AI Overviews and AI Mode are controlled by the normal Search controls, not by Google-Extended. Second, it is one switch for two things: blocking it to opt out of Gemini training also removes the grounding route. Google's 30 October 2026 Search Central post says websites that specifically allow Google-Extended are eligible to appear in responses and receive clicks via grounding in Gemini Apps. The same post says Google plans to support a shared, cross-provider AI training opt-out in robots.txt, based on the IETF AIPREF work, alongside other providers in 2027.

So for Gemini, ordinary Google SEO is the main work: indexing, crawlable text, and pages that rank for the question.

How to show up in Claude answers

Claude cites web pages when it searches, so allow Claude-SearchBot and Claude-User; ClaudeBot is the training crawler and is a separate decision. Anthropic's crawler help article describes all three:

• Claude-SearchBot crawls to improve the relevance and accuracy of search responses. Blocking it may reduce your visibility and accuracy in user search results.
• Claude-User fetches pages when a Claude user asks a question that needs them. Blocking it may reduce your visibility for user-directed web search.
• ClaudeBot collects content that may contribute to model training. Blocking it opts future content out of training.

Anthropic says its bots honour robots.txt and the non-standard Crawl-delay, and advises against blocking its IP addresses, because then the bots cannot read your robots.txt. Its web search help says every response that uses web search includes citations, and that in the newer Claude experience Claude searches the web when it helps rather than waiting for a toggle.

How to show up in Microsoft Copilot

Microsoft reports Copilot citations inside Bing Webmaster Tools, so the route in is Bing: get indexed, keep sitemaps and IndexNow current, and watch the AI Performance report. Microsoft's AI Performance announcement (10 February 2026) says the report covers Microsoft Copilot, AI-generated summaries in Bing and select partner integrations.

If you have not set up Bing yet, our Bing Webmaster Tools guide walks through the import from Search Console, which takes a few minutes. Check that bingbot is not blocked by robots.txt or your CDN.

What kind of content gets cited in AI answers?

The pages that get cited answer a specific question plainly, back it with facts, and say something other pages do not. The engines rarely publish ranking factors, but Microsoft and Google have each said what they look for, and their advice lines up.

• Microsoft, for Copilot. The AI Performance post suggests structuring pages with clear headings, tables and FAQ sections; supporting claims with examples, data and cited sources; keeping content current; describing the same entities consistently in text, images and video; using IndexNow when content changes; and keeping Bing Places details current for local businesses.
• Google, for its AI features. Google's generative AI optimization guide puts "non-commodity" content first: first-hand experience and expert views rather than generic tips. It also warns that creating a separate page for every variation of a question to manipulate AI answers breaks its scaled content abuse policy.

In practice that means a page per real question, not per phrasing. Here is a worked example for a small accounting firm that wants to be named for "best bookkeeper for Shopify stores in Toronto":

1. Answer the question in the first paragraph

   "We do bookkeeping for Shopify stores in Toronto from $X a month, including sales tax filing and payout reconciliation." An engine can quote that sentence as it is.

2. Put the facts in plain text

   Prices, what is included, turnaround, software supported, service area. A table works well. Engines cannot cite a price that only appears in a PDF or an image.

3. Add what only you know

   A short case from your own work, the three mistakes you see most in Shopify payouts, a checklist you actually use. This is the non-commodity part.

4. Keep the same facts everywhere

   Your Business Profile, Bing Places, directory listings and site should all agree on name, prices and services, so an engine is not choosing between two versions of you.

What not to do: Google's guide lists chunking content into tiny pieces, rewriting pages just for AI, chasing inauthentic mentions and special AI markup as unnecessary for its AI features. An llms.txt file is optional; Google Search does not use it.

How do you measure AI search visibility?

No AI assistant gives site owners a full report, so you measure AI visibility by combining manual prompt checks, Bing's AI Performance report, Search Console, and a tracked set of questions. Each covers a different piece.

1. Write 10 to 20 real buyer questions

   Use the words customers use in calls and emails, not your keywords. Mix "best X for Y", "X vs Y" and "how do I" questions. Our guide to checking if ChatGPT mentions your brand has the full manual method and a spreadsheet layout.

2. Record the baseline

   For each engine and question: were you named, in what order, and which domains were cited. Keep the answer text.

3. Read Bing's AI Performance report

   Microsoft suggests using it to see which pages are already cited and to find indexed pages that are rarely cited. Those rarely cited pages are your first rewrite list.

4. Repeat on a schedule

   One answer is an anecdote. The same questions asked weekly show a trend you can act on.

How MrKeyword tracks AI visibility

MrKeyword's AI visibility feature asks ChatGPT, Gemini, Claude and Perplexity your chosen questions every week and records whether each answer named you, in what position, which competitors it named and which sites it cited. You can re-run a question on demand once every 24 hours. The question allowance is one pool for the whole workspace:

A one-time $5 pack adds 25 questions to the pool for 12 months. All limits are on the pricing page. The Site Audit, on every plan, flags robots.txt rules that block AI crawlers and content that needs JavaScript to appear, which covers the access half of this guide.

What MrKeyword does not do

MrKeyword measures and audits; it cannot make any AI engine recommend you, and it does not see exactly what an individual user was shown. Specifically:

• It uses the engines' APIs, not the consumer apps. In MrKeyword's checks, ChatGPT, Gemini and Claude answer without web search, so they measure brand memory; Perplexity searches the web, so it measures current web evidence. The consumer apps can search, so a person may see a different answer.
• It does not measure Copilot, Google AI Overviews or AI Mode. Use Bing's AI Performance report and Search Console for those. MrKeyword does not import either report.
• No "AI score" or AI traffic estimate. Every number is a count of stored answers you can read.
• No guaranteed placement. Nobody outside these companies controls their answers, and any tool that promises a ChatGPT ranking is promising something it cannot deliver.

Questions and answers