How to Rank in ChatGPT: What Actually Moves Citations

"Ranking in ChatGPT" means being one of the two or three products the model names when someone asks about your category — and there is no rank tracker for it, because there is no ranked list. ChatGPT's answers come from retrieval over crawlable sources plus what the model already knows, so the levers that move citations are the ones that change what retrieval finds and what the model trusts. Here are the seven, in the order they usually pay off, and the check that tells you which one you are missing.

1. Let the crawler in

If ChatGPT's crawler cannot fetch your pages, nothing else matters. Check your robots.txt for the AI user-agents and confirm your key pages return 200 to them; the full rule set (including which agents matter and what llms.txt adds) is in our AI crawler guide. A site that blocks AI crawlers "to save bandwidth" has silently opted out of this channel.

2. Give it one sentence to quote

Models quote definitions. The single highest-leverage edit on most product sites is a first paragraph that says, in plain words, what the product is, for whom, and what it does — the same sentence everywhere: homepage, directory listings, review profiles. When versions disagree, retrieval surfaces the disagreement and the model hedges or skips you.

3. Publish the question pages

For each category question you want ("best X for Y", "X vs Z", "how much does X cost"), have one page that answers it in the first two sentences and backs it with a table or list. ChatGPT's retrieval favors pages whose structure matches the question's structure. This is the same asset the AI SEO playbook builds in step 2.

4. Get named by other domains

A mention inside a third-party roundup or comparison page carries more weight than anything on your own site, because it is the model's equivalent of a reference letter. Two practical pools: "best X" roundups and "X alternatives" pages in your category — both are citation magnets. Getting into them is outreach plus a listing footprint on credible directories (aat.ee's network is built for exactly this: permanent dofollow listings on indexed domains).

5. Keep your facts boring

Pricing, limits, platform support, founding year: the unglamorous facts are what assistants recite, and reciting them wrong is how products get "corrected" in answers. One canonical fact sheet, versioned like code, reused verbatim.

6. Stay fresh where freshness is visible

Changelog pages, updated comparison tables, and honest "last updated" dates matter because retrieval recency is one of the tie-breakers between two otherwise equal sources. A 2024-stamped page losing to a 2026-stamped competitor is the most common silent loss we see.

7. Measure, then fix the weakest link

Ask your ten category questions from a logged-out session, weekly, per assistant, and log which products get named and from which sources (ChatGPT shows citations on search answers). The gap analysis tells you which lever above is missing: not crawled (1), no quotable sentence (2), no question page (3), no third-party mentions (4), contradictory facts (5), stale pages (6). For the Search Console side of measurement, see tracking GEO with Search Console.

What does not work

Keyword repetition inside your own pages, hidden "prompt injection" text (it gets you flagged, not cited), and paid "guaranteed ChatGPT ranking" services — there is no auction and no backdoor, only retrieval and trust. The one exception worth naming: schema.org markup does not directly create citations, but it reduces extraction errors, which is why we still recommend it in schema.org for AI answers.

FAQ

Does ChatGPT use Google's index? No — it retrieves from its own crawl and browsing stack, which is why robots.txt rules for AI agents matter independently of Googlebot.

How many sources does an answer draw on? Search-mode answers cite a handful of pages; conversational answers blend retrieval with training knowledge. Both reward the same levers: credible, consistent, crawlable facts.

Can a brand-new product get cited? Yes, on narrow questions, faster than on head terms. Freshness and specificity favor newcomers; authority accumulates later.

Is this different for Perplexity or Gemini? The levers are the same; the weights differ (Perplexity leans harder on live retrieval and visible citations). See which assistant cites you for the per-engine check.

Start where retrieval looks: a permanent, dofollow listing on indexed directories puts your facts in the citation pool. Submit your product on aat.ee or compare the networks.