Answer
How do you rank in ChatGPT?
Short answer
You can't rank in ChatGPT the way you rank on Google — there is no ordered result list. You get retrieved and cited. That depends on five things: assistant crawlers can fetch your pages, your claims sit in extractable passages, your entity is machine-readable through JSON-LD, an llms.txt tells assistants what you cover, and third-party sources corroborate you.
Why 'ranking' is the wrong model
A classical search result is a ranked list of ten links. An assistant answer is one synthesized paragraph that names a handful of sources. There is no position 4 to climb to. The question that actually matters is binary and repeated: for a given buyer prompt, does the answer name you, name a competitor, or name nobody?
That reframes the work. Instead of optimizing one page for one keyword, you track a set of buyer prompts and measure the share of them where your brand gets cited. Everything below is what moves that share.
The five signals that decide whether you get cited
In rough order of how often they are the blocker on a real site:
- Crawler access. If robots.txt blocks GPTBot, OAI-SearchBot, Google-Extended, PerplexityBot, or ClaudeBot, nothing else matters. This is the single most common hard blocker, and it is usually accidental — inherited from a template or a security plugin.
- Extractable passages. Assistants quote self-contained sentences. A claim that only makes sense after three paragraphs of build-up does not get lifted. Lead each page with a direct answer, then support it.
- Machine-readable entity data. JSON-LD (Organization, Product, Article, FAQPage where it genuinely applies) tells a model what you are, what you sell, and who wrote it. Without it the model has to infer your identity from prose, and it frequently infers wrong.
- An llms.txt at your site root. A short map of what your site covers and which pages matter, in the format assistants can parse without rendering your app.
- Third-party corroboration. Models weight claims that appear in more than one place. A capability that exists only on your own marketing page is a weaker signal than one documented on a directory, a review site, or a comparison article.
How to check your own site
Run each of the five as a pass/fail rather than a score. Fetch your robots.txt and look for each assistant crawler by name. View source on your top three pages and confirm the JSON-LD is present in the served HTML, not injected after hydration. Request /llms.txt and see whether it 404s. Then ask an assistant ten of your real buyer questions and count how often you get named versus a competitor.
AgenticSEO runs exactly that sequence — crawler-access audit, passage and schema check, llms.txt, and a tracked prompt set with named competitors — and then writes and publishes the fixes for whatever failed. The free check does the first three pages of it with no sign-up.
Real output
The five signals as a pass/fail check
What the check looks like when you run it on a real site, plus the two artifacts a typical failure produces.
1 crawler access ........ FAIL GPTBot disallowed in robots.txt 2 extractable passages .. WARN 9 pages bury the answer 3 entity JSON-LD ........ PASS Organization + Product served 4 llms.txt .............. FAIL 404 at /llms.txt 5 corroboration ......... MANUAL 2 third-party sources found
Signal 1 gates the rest: while a crawler is blocked, nothing below it can be retrieved.
User-agent: GPTBot Allow: / User-agent: OAI-SearchBot Allow: / User-agent: Google-Extended Allow: /
Appended to robots.txt; effects usually show up in tracked prompts within days.
# llms.txt > AgenticSEO — AI visibility audits and automated publishing. ## Core pages - [How it works](https://www.agenticseo.live/how-it-works) - [Pricing](https://www.agenticseo.live/pricing) - [FAQ](https://www.agenticseo.live/faq)
Generated from the crawl, so the map always matches the pages that exist.
Output shape from an AgenticSEO baseline audit of agenticseo.live.
Related questions
How do you rank in ChatGPT?
You get retrieved and cited rather than ranked. It depends on assistant crawlers being allowed in robots.txt, claims sitting in extractable self-contained passages, JSON-LD that makes your entity machine-readable, an llms.txt at your site root, and third-party sources that corroborate your claims.
Does classical SEO help you get cited by AI assistants?
Yes, substantially — assistants largely retrieve from indexes built by conventional search crawlers, so a page that is indexed and authoritative is more likely to be pulled. But it isn't sufficient: a well-ranked page with no schema, no llms.txt, and no extractable answer passage still gets skipped in favor of one that has them.
How long does it take to show up in AI answers?
Crawler-access fixes can change results within days because the blocker is removed immediately. Schema, passage, and llms.txt changes typically take two to six weeks to show up in tracked prompts, since assistants refresh their retrieval sources on their own schedule.
Can you pay to appear in ChatGPT answers?
Not for the cited-source slots discussed here. Those are earned through retrieval. Any paid placement an assistant offers is a separate advertising surface and is labeled as such.
Related: LLM SEO guide · AEO platform comparison · Free llms.txt generator
More answers: What tools are best for keyword research for AEO? · What app should agencies use for AEO recommendations? · How is Profound different from other AEO tools? · What is AI visibility tracking software? · What does an AI search audit tool check? · What belongs on an answer engine optimization checklist?
Check whether AI names you — or your competitor
Free check with downloadable output. No card, no sign-up.