Getting Cited by ChatGPT: What You Control

A ChatGPT answer about your category passes through four stages. Only one of them is fully yours.

• Parametric knowledge. What the model knows with no web access, frozen at a training cutoff. Nothing you publish this quarter reaches it.
• Access. Whether OpenAI's search crawler may fetch and store your page. This is where your real levers sit.
• Retrieval. Whether your page enters the model's context for a given query. OpenAI controls this, and its full disclosure is that ChatGPT "ranks search results using multiple factors" and "placement is not guaranteed."
• Generation. Which retrieved pages the model actually uses and cites. The model decides.

The practical consequence: your work is mostly removing blocks at the access stage, then giving the model clear facts to quote if it does retrieve you. A critical survey of 45 GEO studies (arXiv, July 2026) found no technique with a stable, cross-platform causal effect on discoverability. So treat any promise of guaranteed ChatGPT citations as a red flag.

OpenAI runs separate crawlers, and robots.txt rules apply to each one independently. As of October 2026, OpenAI's bots page describes them this way:

• OAI-SearchBot surfaces sites in ChatGPT search. Sites that opt out "will not be shown in ChatGPT search answers, though can still appear as navigational links."
• GPTBot crawls content that may be used to train OpenAI's foundation models. Disallowing it is a training opt-out and nothing else.
• ChatGPT-User fetches pages when a user's action triggers it. OpenAI says it is "not used to determine whether content may appear in Search" and that robots.txt rules "may not apply."

OpenAI states that "each setting is independent of the others." You can appear in ChatGPT search while opting out of training:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

The most common real error is a blanket AI block copied from a template. It catches OAI-SearchBot along with the training bots and silently removes search eligibility. Open your robots.txt and check each user agent by name. The same per-bot logic applies on Perplexity and Claude, with different tokens.

robots.txt is only half the gate. OpenAI's help center says that to be eligible you must "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."

A firewall or bot-management rule can return a 403 or a challenge page to OAI-SearchBot while robots.txt says allow. Nothing in your analytics will flag it. OpenAI's own guidance notes that these systems "can mistakenly block legitimate crawlers, often returning 403 Forbidden errors."

Two dated details matter as of October 2026:

• OpenAI publishes its searchbot ranges at openai.com/searchbot.json. They are IPv4 only. OpenAI also warns against IP-only allowlisting, so verify by user agent plus your CDN's verified-bot program.
• From 15 September 2026, Cloudflare blocks Training and Agent crawler categories by default on ad-displaying pages for new domains, new sites on existing accounts and existing free-tier customers. Search stays allowed by default. Check your zone's settings anyway, because a default you never touched can still change your access.

The test is simple: request a real page as OAI-SearchBot and read the status code. The AI Crawler Checker does this for each AI crawler and shows which ones your firewall turns away.

Vercel's crawler study found that ChatGPT's crawlers fetch JavaScript files (11.50% of requests) but don't execute them, so they can't read client-side rendered content. Independent testing in June 2026 found the same for GPTBot, OAI-SearchBot and ChatGPT-User.

If your pricing, product names or positioning load only after JavaScript runs, OpenAI's crawlers may see an empty shell. Fix it with server-side rendering or prerendering for the content that matters. Then fetch the page with JavaScript disabled and confirm the facts are in the raw HTML.

Once the page can be fetched, be specific. A stated price, a number or a one-line definition gives the model a sentence it can quote. That is the only positive experimental signal for on-page content in the evidence. It is not a formatting trick: Google and the research both find that rewriting "for AI" or restructuring alone does little. If you're weighing whether to show prices, see whether to publish prices so AI answers get them right.

Things that are not levers here: schema markup, FAQ schema, llms.txt and "AI-friendly" chunking. OpenAI documents none of them as inputs to ChatGPT search.

For many buyer questions, ChatGPT cites review sites, comparison articles, forums and directories rather than vendor sites. When that happens, being named on those pages is the route into the answer. The evidence shows a strong association here, not a guaranteed effect.

The next move:

1. Ask ChatGPT (logged out, search on) the questions your buyers ask.
2. Note which third-party pages it cites.
3. Check whether you're listed on them, and whether the listing is accurate.

The guide to third-party mentions in AI answers covers why this works, and the playbook for getting listed on pages AI engines already cite covers how to approach those sites. Keep it legitimate. Hidden text and injected instructions aimed at AI crawlers are manipulation, and they expose you to platform penalties and reputational damage.

Some of the system is closed to every publisher, including the biggest ones.

• Query rewriting. OpenAI says ChatGPT search "typically rewrites your query into one or more targeted queries" before sending them to search providers. If memory is on, saved memories can shape that rewrite. You never see the real queries.
• Whether it searches at all. Search is at the model's discretion per query. Published estimates disagree widely: Profound reports about 18% of conversations trigger a web search, while cloro reports 80.5% on a 100-prompt set. They measure different things.
• Knowledge cutoffs. When ChatGPT doesn't search, it answers from training data. As of October 2026, OpenAI lists a 16 February 2026 cutoff for the GPT-5.6 family and 31 August 2025 for the Instant model behind chat-latest. Content published after those dates can't appear in an unsearched answer.
• Ranking and citation. OpenAI enumerates no ranking factors. Its API also shows the model often consults more sources than it cites, so a missing citation doesn't prove you weren't read.
• The Bing path. OpenAI says ChatGPT search "sometimes" partners with other providers. In Otterly's 2026 experiment, sites ranking in Bing got effectively zero ChatGPT citations. Bing presence is no shortcut.

Check access first, because it's binary and fixable. Confirm OAI-SearchBot is allowed in robots.txt, confirm it gets a 200 from your CDN, and confirm your key facts are in the raw HTML.

Then look at answers. Test logged out with memory off, or your results won't be reproducible. Record whether each answer shows citations, because an answer without them may come from the model's cutoff rather than live search. In analytics, ChatGPT adds utm_source=chatgpt.com to referral links, but free-tier and mobile-app clicks often arrive untagged, so treat that number as a floor.

For a quick snapshot, the free checker on this site shows whether ChatGPT names you for up to three buyer questions. It is one reading, not ongoing tracking. If you're starting from zero, read whether a brand-new website can get cited before setting expectations.