ChatGPT Ranking Factors: What OpenAI Actually Discloses

OpenAI's entire public statement on ranking, from its help page on ChatGPT search, is this: ChatGPT "ranks search results using multiple factors intended to help users find relevant, reliable information. Placement is not guaranteed."

That is the full disclosure as of October 2026. No factor is named. None is weighted or ordered. OpenAI publishes no position data and no ranking report for site owners.

So any list of "ChatGPT ranking factors" you find elsewhere is one of three things: a guess, a correlation study, or a lab experiment on content already loaded into a model. Correlations in this field are heavily confounded by brand size, which is why it pays to read the difference between correlation and causation in AI search studies before acting on one.

What OpenAI does document is narrower and more useful. It tells you which crawler decides search eligibility, what your network has to allow, and how ChatGPT turns a prompt into searches. Those are the parts of this page you can act on.

A ChatGPT answer passes through up to four stages. Knowing which one you are looking at tells you whether anything you do can reach it.

Parametric: no search at all

Many answers never touch the web. ChatGPT decides per prompt whether to search, so the model often answers from what it learned in training. OpenAI's model docs list a knowledge cutoff of February 16, 2026 for the GPT-5.6 family and August 31, 2025 for the Instant path (chat-latest). Nothing you publish this quarter changes a retrieval-off answer.

Access: can OpenAI's crawler fetch you

This is index-time eligibility. It depends on robots.txt, your firewall or CDN, and whether your content is in the HTML. You control almost all of it.

Retrieval: does your page enter the answer's context

When ChatGPT searches, OpenAI says it "typically rewrites your query into one or more targeted queries" and sends them to search providers, which it says it "sometimes" partners with. If memory is on, saved memories can shape that rewrite. You never see these queries.

Generation: which retrieved pages get cited

Retrieved and cited are different sets. OpenAI's API documentation says the number of sources consulted "is often greater than the number of citations." A page can inform an answer without being linked, and a citation count is a noisy proxy for influence.

What it means: the only stage with documented, checkable requirements is access. Start there.

OpenAI runs three crawlers, and they do different jobs. Confusing them is the most common real-world error.

OpenAI states that "each setting is independent of the others." A site can allow OAI-SearchBot to appear in search while disallowing GPTBot to opt out of training. This robots.txt does exactly that:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

Allowing OAI-SearchBot buys eligibility. It does not buy retrieval, a citation or a position. Blocking it, deliberately or through a blanket "block all AI" rule, removes you from ChatGPT search answers.

If your logs show one OpenAI bot and not another, see why a server log can show GPTBot but not OAI-SearchBot.

robots.txt is only half of OpenAI's eligibility rule. Its help page says: "allow OAI-Searchbot to crawl the site and confirm that the website host or content delivery network allows traffic from OpenAI's published searchbot IP addresses." The word is "and".

OpenAI publishes those ranges at openai.com/searchbot.json. When the evidence behind this page was verified in August 2026, the file listed IPv4 prefixes and no IPv6 prefixes. Check the current file rather than copying an old list.

Where this breaks

• Bot mitigation returns 403. OpenAI's own guidance notes these systems "can mistakenly block legitimate crawlers, often returning 403 Forbidden errors." Your robots.txt says allow, your WAF says no, and no dashboard tells you.
• JavaScript challenges and CAPTCHAs. An automated crawler can't pass them, so it gets a challenge page instead of your content.
• CDN defaults change under you. From September 15, 2026, Cloudflare blocks Training and Agent crawler categories by default on ad-displaying pages for new domains, new sites and existing free-tier customers, while Search stays allowed by default. Check which categories your zone actually blocks.

IP lists alone are a weak allowlist. Pair user-agent checks with the CDN's verified-bot programme.

Rendering

Vercel's crawler study found that ChatGPT's crawlers fetch JavaScript files but "don't execute them. They can't read client-side rendered content." Independent retesting in June 2026 confirmed this for GPTBot, OAI-SearchBot and ChatGPT-User. If your pricing, product names or key claims appear only after JavaScript runs, OpenAI's crawlers see a page without them.

To see what each crawler actually gets back, the AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away.

Be plain with yourself and your team about this split. It decides where budget goes.

In your control

• Per-bot robots.txt rules for OAI-SearchBot and GPTBot.
• Whether your host, CDN and WAF let OpenAI's crawlers through.
• Whether the content that matters is in the server-rendered HTML.
• Stating specific facts (a price, a number, a one-line definition) on a page that can be fetched. That gives the engine a sentence it can quote, and it is the only positive experimental signal for on-page content.
• Keeping those facts consistent across your pricing page, about page and footer.
• Being named on the third-party sites ChatGPT already cites for your topic. The evidence shows a strong association here, not a promised effect.

Not in your control

• Whether ChatGPT searches for a given prompt.
• The rewritten queries it sends, and how memory shapes them.
• Which search providers it uses for a query.
• Where your page sits in the model's context. The C-SEO Bench study found context position mattered far more than content rewrites, but that position was set by researchers, not earned.
• What the model learned before its knowledge cutoff.
• Which retrieved pages end up cited.

No one can promise you a ChatGPT citation or a position. What you can do is make sure nothing on your side blocks you.

Work from the cheapest, most decisive checks to the slowest.

1. Read your robots.txt per bot. Confirm OAI-SearchBot is allowed. Decide on GPTBot separately, as a training choice.
2. Request a real page as OAI-SearchBot through your live CDN. A 200 with your content passes. A 403, a 429 or a challenge page fails. Repeat for ChatGPT-User.
3. Diff raw HTML against the rendered page. List every price, product name and claim that exists only after JavaScript runs. Move those into the server-rendered HTML.
4. State your key facts plainly and once. One clear sentence for your price, what you do and who it's for. Fix any page where the same fact differs.
5. Find out who ChatGPT cites for your buyer questions. If it names review sites, directories or comparison pages, being listed accurately there is the route into those answers. The free checker on this site shows whether ChatGPT names you for up to three buyer questions, as a snapshot.
6. Test cleanly. Run checks logged out with memory off, and note whether the answer showed citations. Without citations, you can't tell a search answer from a training-data answer.

Steps 1 to 3 are fixes with documented consequences. Steps 4 to 6 improve your odds without guarantees. Say so when you report results.

These come up constantly, and each one is contradicted by OpenAI's own documents or the best available studies.

• "Allow GPTBot so ChatGPT can find you." GPTBot governs training only. OpenAI asserts no search benefit from it.
• "Blocking GPTBot removes you from ChatGPT." It doesn't. OAI-SearchBot is the search control, and the settings are independent.
• Treating robots.txt as the whole gate. OpenAI's rule includes network reachability. A 403 at the CDN is a silent, complete failure.
• Assuming a Bing or Google ranking carries over. OpenAI's provider language is hedged. In Otterly's 2,000 AI blogs test, ranking in Bing produced zero ChatGPT citations even while ChatGPT's crawlers visited, though that was one pair of sites. For the Google side, see whether you need to rank on Google to show up in ChatGPT and how organic ranking relates to AI citations.
• Rewriting pages in an "AI-friendly" style. C-SEO Bench found most conversational SEO methods "largely ineffective" and often harmful to ranking in the model's context.
• Adding schema or an llms.txt for ChatGPT. OpenAI documents neither as an input. Ahrefs' matched test found schema produced no major citation uplift on any platform.
• Tracking a "ChatGPT rank". No ground truth exists to check it against. Read why there's no AI rank to track and be wary of leaderboards that rank brands.
• Copying one engine's rules to another. Perplexity uses different crawlers and controls; its visibility factors need their own checks.