The short answer
To get cited by ChatGPT, a page must first be findable by OpenAI's search crawler, then be worth quoting once it is found. OpenAI documents the access step you control: allow OAI-SearchBot, its search crawler, in your robots.txt file and let requests from its published IP addresses through. Sites that opt out of that crawler will not be shown in ChatGPT search answers, though they can still appear as navigational links. Beyond access, OpenAI's crawler documentation says nothing about schema, authorship or ranking, so treat those as habits, not rules. Start with access, because good writing cannot help a page the crawler cannot reach, and a robots.txt change takes about 24 hours to register. Then answer one question per section. Then measure by asking ChatGPT your buyers' questions every month and logging which pages it names.
On this page
Which OpenAI crawler decides whether ChatGPT search can use your page?
OAI-SearchBot. OpenAI says that crawler surfaces websites in ChatGPT's search features, and that blocked sites can still appear as navigational links. The other OpenAI bots do different jobs. Only OAI-SearchBot and GPTBot are managed through robots.txt, and OpenAI says those two settings are independent.
OpenAI lists four bots in its overview of OpenAI crawlers. Only one of them governs search.
| Bot | What OpenAI says it does | What to do |
|---|---|---|
| OAI-SearchBot | Surfaces websites in ChatGPT's search features | Allow it |
| GPTBot | Crawls content that may be used to train OpenAI's models | Your choice, and separate from search |
| ChatGPT-User | May visit a page when a person asks ChatGPT a question. OpenAI says robots.txt rules may not apply, and that this bot does not decide whether content appears in Search | Do not use robots.txt to manage search with it |
| OAI-AdsBot | Checks landing pages for ads submitted on ChatGPT | Only matters if you advertise there |
OpenAI states that these settings are independent. A site can allow OAI-SearchBot to appear in search results and disallow GPTBot to signal that its content should not be used to train OpenAI's models. A single rule that blocks every AI bot also blocks the search crawler, so it is worth checking that your file does not carry one.
Two more details from the same page. OpenAI says a robots.txt change can take about 24 hours to take effect in search. And it recommends allowing requests from its published IP ranges, not only editing robots.txt. The search crawler's ranges are listed at openai.com/searchbot.json.
What did we find when we tested seven sites?
None of the seven blocked an OpenAI crawler by name. For these sites, robots.txt was not what stood between them and a ChatGPT citation.
On 17 October 2026 we tested seven sites we have built or managed, spread across four hosting setups. For each one we read robots.txt, then fetched the homepage four times: as a normal browser, then as OAI-SearchBot, GPTBot and ChatGPT-User. We compared the status code and the page size.
All seven answered with a normal page. Six returned exactly the same size to every user agent. The seventh differed by seven bytes of page content; nothing was blocked. No robots.txt named an AI bot. Only one site sat behind Cloudflare, and that was the layer Shopify runs for its own stores, so its AI crawler settings were never ours to change.
This test has a limit. Every request came from one computer outside OpenAI's network. Hosts and security layers can recognise a bot by its IP address and treat the real crawler differently from a copy that only borrows its name. A pass proves that no rule blocks by name. It does not prove that OpenAI's crawler gets through. Your host's firewall settings or access logs are the real proof.
How do you check that ChatGPT can find a page?
Check five things in order: robots.txt, your host's bot protection, the page's own indexability, whether the answer sits in plain HTML, and the 24-hour delay. This takes about 20 minutes.
- Read your robots.txt. Open yourdomain.com/robots.txt. Look for any line naming OAI-SearchBot, and for a blanket
Disallow: /that applies to every bot. A rule written for one purpose can catch bots it was never meant for. - Check your host's bot protection. Firewall and "block AI bots" switches can stop the crawler even when robots.txt is open. Cloudflare sites have a dashboard control for this, and Vercel has one in its project Firewall. Check what your own host offers, and look for anything set to deny or challenge AI bots.
- Check the page itself. It should be live, free of a noindex tag, listed in your sitemap, and linked from at least one other page on your site. These are standard indexability checks, not OpenAI requirements.
- Read the page as plain HTML. Put the answer in text that loads without scripts. OpenAI's crawler page does not say how its crawler treats scripted content, so the safe version is text that needs no script to appear.
- Wait about 24 hours after any robots.txt change before you judge the result.
Once these pass, the crawler can reach the page. What happens next is the part nobody can guarantee.
Is being found enough to be cited?
No. In a joint post with Cloudflare, OpenAI says it evaluates and ranks found pages on how relevant and helpful they are. Being found only gets a page that far.
On 15 October 2026, Cloudflare and OpenAI published early results from a pilot that began in July 2026. It covered about 1% of Cloudflare's network. According to the Cloudflare and OpenAI post, the pilot helped OpenAI discover and index hundreds of millions of new web pages. The post says OpenAI then independently evaluates and ranks how those pages appear, based on how relevant and helpful they are. It also says "Finding a URL is only the start." In context, that line is about weeding out dead addresses and duplicate copies before a crawl.
That is a stated principle, not a formula. No one outside OpenAI can give you the formula, and this article does not claim to.
Google's AI Overviews are a separate engine with their own selection, which we cover in how AI Overviews choose which brands to cite. If your worry is clicks lost to answers on the results page, see our zero-click survival plan.
What makes a found page worth quoting?
Pages that say something specific and checkable are easier to quote than pages that restate what ten others already say. These habits are practice, not OpenAI rules.
- Open each section with the answer. Put the direct answer in the first two sentences, then explain. A system looking for a passage to lift has less to untangle.
- Add something only you can add. A test you ran, a number you measured, a client result you can show. This article leads with our own crawler test for that reason. A page that repeats the top results gives a system nothing new to say.
- Put a real name on it. A byline, a short specific bio, and the same spelling of the name everywhere it appears. OpenAI's crawler page does not mention authorship. Google's guidance on helpful content notes that it helps people judge content when it is clear who created it. That is about Google Search, not a ChatGPT rule, but it is the checkable version of "who said this".
- Keep schema, but do not expect it to do the work. We add structured data to our articles because it states plainly what a page is. OpenAI's crawler page does not mention it, and we have seen no test showing it moves ChatGPT citations. Treat it as tidy labelling.
- Make the same facts true elsewhere. Reviews, press, directories and partner pages that repeat your name and description support you off your own site. This is sound reputation work. It is not a documented ChatGPT signal.
How do you know whether any of it worked?
Ask ChatGPT your buyers' real questions once a month and log what it names. One answer proves nothing, because output varies between runs. Read the pattern over a quarter.
Search Console shows you clicks and rankings. We know of no equivalent report for ChatGPT citations, so the check is manual. Write down ten to twelve questions a buyer would type before choosing a business like yours. Run them monthly. For each answer log four things: the date, whether your brand is named, which page is linked, and which sources are cited instead.
The fourth column is the useful one. The pages ChatGPT names in your place are your real competition for that question. Read them. Ask what they say that your page does not.
Should you do this yourself or hand it over?
You can do all of it yourself. The question is whether the time exists in your week.
The access check takes about 20 minutes and needs no developer. The monthly log takes about an hour each month, every month. None of it is beyond a careful non-developer. If the hour exists, do it yourself and pay us nothing. Our free AI visibility check is a separate, quick test: it runs Google's agentic browsing audit on a page you choose, plus a probe for an llms.txt file, with no email needed. It tells you whether AI agents can use your page. It does not tell you whether ChatGPT will cite it, and we do not promise that anyone will appear in ChatGPT answers.
If you would hand it over anyway, that is our SEO and AI visibility service. Tell us about your business and we will say honestly whether we are the right fit.
Frequently asked.
01Do I need to allow GPTBot to be cited by ChatGPT?
No. OpenAI says GPTBot and OAI-SearchBot are separate settings. GPTBot relates to training. OAI-SearchBot relates to search. You can allow the search crawler and disallow the training crawler at the same time. Disallowing GPTBot signals that your content should not be used to train OpenAI's models. It does not, by itself, change what the search crawler may do.
02How long after I change robots.txt will ChatGPT notice?
OpenAI says it can take about 24 hours for its search systems to adjust after a robots.txt update. Judge the change after that, not before. Note the date you edited the file so you can match it to later results. Firewall changes at your host may follow their own timing, so check those separately.
03Can I pay or submit my site to be cited?
We found no submission form or paid placement for organic citations in OpenAI's crawler documentation. Ads shown in ChatGPT are a separate product, checked by a separate bot. The lever you control for organic citation is access, plus a page worth quoting.
04Does schema markup get a page cited by ChatGPT?
We know of no published evidence that it does. OpenAI's crawler page does not mention schema. It is still worth adding, because it states plainly what the page is and who wrote it. Do not expect it to decide a citation alone.
05Why would a page be crawlable and still not get cited?
Access only gets the page considered. OpenAI says, in a joint post with Cloudflare, that it ranks found pages on how relevant and helpful they are. A page that repeats what other pages say gives a system little reason to choose it.
06Can a small brand earn a ChatGPT citation?
It is possible on a narrow question where your answer is genuinely specific. We have no study that gives odds. Pick questions where you hold first-hand information and answer them better than anyone. List those questions before you start; they are the ones to test first in your monthly log.