How to Rank in ChatGPT: What's Actually Possible

"Ranking in ChatGPT" borrows a word from Google that doesn't fit. A ChatGPT answer is generated fresh for each conversation. Two people asking the same question can get different brands, in a different order, with different sources.

OpenAI's own disclosure is short. Its help page on ChatGPT search says results are ranked "using multiple factors intended to help users find relevant, reliable information" and that "placement is not guaranteed." That is the whole list. We cover what it does and doesn't tell you in ChatGPT ranking factors.

So the useful question isn't "how do I rank." It's "what stops ChatGPT from finding and naming me, and can I remove it." That question has concrete answers. Any tool selling a ChatGPT position is reporting a number with nothing to check it against, which is why AI rank checkers have no rank to check.

ChatGPT answers from one of two places, and it decides which.

Without search: what the model already knows

If ChatGPT doesn't search, it answers from its training data. That knowledge is frozen at a cutoff date, and different ChatGPT models carry different cutoffs. Nothing you publish this quarter reaches these answers. The only route in is future training, which takes a full model generation. See how a brand ends up in training data for that slower path.

With search: what it retrieves today

When ChatGPT searches, it doesn't send your customer's words straight to a search engine. OpenAI says ChatGPT search "typically rewrites your query into one or more targeted queries" and sends those to its search providers. If memory is on, saved memories can shape that rewrite. You can't see the real queries, and you can't optimize for an input you can't observe.

Retrieval happens in stages. First a crawler must be allowed to fetch your page (access). Then the page must be pulled in for a given query (retrieval). Then the model chooses what to use and cite (generation). You have real control over the first stage. You have partial influence over the second. The third belongs to the model.

One more wrinkle: the pages ChatGPT retrieves and the pages it cites are different sets. OpenAI's API documentation notes that the number of sources consulted is often greater than the number of citations shown.

OpenAI runs separate crawlers, and each robots.txt setting is independent. Mixing them up is the most common access mistake.

OpenAI's bots page states "each setting is independent of the others." A site can allow search and refuse training at the same time:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

That configuration is legitimate. A blanket rule that blocks every AI user agent is not neutral: it silently removes your ChatGPT search eligibility along with training.

In your control

• robots.txt permission for OAI-SearchBot. This is the stated gate for ChatGPT search.
• Network reachability. OpenAI's help centre says eligibility means allowing OAI-SearchBot and confirming your host or CDN allows traffic from OpenAI's published search bot IP addresses. A firewall can return 403 to a crawler your robots.txt welcomes, and nothing in your analytics will flag it.
• Server-side rendering. Vercel's crawler study found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they "can't read client-side rendered content." If your pricing or product names load after JavaScript runs, assume ChatGPT's crawlers can't see them.
• Plain, specific facts on your own pages. A price, a plan name, a one-line description of what you do. Stating it on a fetchable page gives the model a sentence it can quote. This is the one on-page signal with positive experimental support, and it promises nothing.
• Consistency. If your pricing page and your about page disagree, you've handed the model two answers.

Not in your control

• Whether ChatGPT decides to search for a given question.
• The queries it rewrites your customer's question into.
• Where your page lands among the candidates it retrieves.
• Which retrieved sources it chooses to cite.
• What older models learned before their cutoff.

A July 2026 critical survey of 45 GEO studies (arXiv 2607.14035) concluded that no reviewed technique shows a stable, cross-platform causal effect on whether content gets discovered. Treat anyone promising a ChatGPT placement accordingly.

When ChatGPT answers "what's the best tool for X," it often draws on other sites: review roundups, comparison articles, forums, industry publications. If those sites name your competitors and not you, your own site can't fix that answer alone.

The evidence shows a strong association between being named on frequently cited sources and appearing in answers. Citation is concentrated: a study of AI search citations (arXiv 2507.05301, data from March to May 2025) found the top 20 news sources accounted for 67.3% of citations for OpenAI models. Big, well-known sources dominate.

What the evidence does not show is a reliable effect you can buy. Brand size confounds almost every study in this area, because large brands get mentioned everywhere and cited everywhere. So the honest framing is this: if engines cite other sites instead of yours, getting accurately named on those sites is the route into the answer, with no promised outcome.

In practice that means earning inclusion on the roundups and directories that already show up for your category, and correcting outdated descriptions of you where they exist. Our page on getting recommended when people ask ChatGPT for the best tools goes deeper. For the obvious shortcut, see whether you can pay to appear in ChatGPT.

One thing never to do: plant hidden text or instructions for AI on your pages or anyone else's. Research on LLM search agents (arXiv 2606.16821) shows manipulated content can sway answers once retrieved. It's also deceptive, it exposes you to platform action, and it invites the same tactic used against you.

Work from the gates outward. Each step removes a block; none guarantees a mention.

1. Read your robots.txt. Confirm OAI-SearchBot is allowed. Decide separately what you want for GPTBot. Remove any blanket AI block you didn't intend.
2. Test reachability, not just permission. Request a key page as OAI-SearchBot and check for a 200, not a 403, a CAPTCHA or a JavaScript challenge. If you're on Cloudflare, check your AI crawler settings: from 15 September 2026 its defaults block Training and Agent crawlers on ad-displaying pages for new domains and existing free-tier customers, while Search stays allowed by default. The AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away.
3. Diff raw HTML against the rendered page. Load your pricing, product and about pages with JavaScript off. Anything missing needs server-side rendering or prerendering.
4. State your core facts plainly. Price, plans, what you do, who it's for, where you operate. One clear sentence each, on pages crawlers can fetch.
5. Fix contradictions across your site. Same price, same plan names, same founding year everywhere.
6. Find the sites ChatGPT cites for your category and work on being accurately named there.
7. Check what ChatGPT actually says, then repeat. If it gets your pricing wrong, start with fixing AI answers that get your pricing wrong.

• Allowing GPTBot to "get found." GPTBot governs training only. OpenAI asserts no search benefit for it anywhere.
• Assuming robots.txt is enough. Permission without network reachability is a silent failure.
• Adding schema or an llms.txt file and expecting citations. Google says structured data isn't required for its generative AI search, and an Ahrefs matched test found AI Overview citations fell 4.6% after schema was added. No engine documents reading llms.txt. Existing markup does no harm; it just isn't a lever for AI answers.
• Rewriting pages in an "AI-friendly" style. Controlled studies of citation-oriented rewrites found them largely ineffective and sometimes harmful.
• Treating Bing rank as a ChatGPT shortcut. In OtterlyAI's 2026 test of two AI-generated blog sites, ranking in Bing produced essentially zero ChatGPT citations even while ChatGPT's crawlers visited. It was one small test, but the shortcut didn't show up. For the Google side, see whether you need Google rankings for ChatGPT.
• Mass-publishing thin pages. The same test saw Google deindex both sites without warning.
• Tracking a "ChatGPT rank." There's no ground truth behind it.

Replace position with questions you can actually answer.

• Is ChatGPT naming you for the buyer questions that matter? Ask them logged out with memory off, because memory changes how queries get rewritten.
• Are the facts right? A wrong price or a dead product name is a fixable problem with a clear owner.
• Is ChatGPT sending visits? OpenAI says ChatGPT adds utm_source=chatgpt.com to referral URLs, so you can segment that traffic in analytics. Treat it as a floor, since not every click arrives tagged.

Repeat the same questions over time and compare like with like. The free checker on this site runs up to three buyer questions and shows whether you're named and how you're described, as a one-time snapshot. For the citation side specifically, see getting cited by ChatGPT.