Quick answer: To get cited by ChatGPT, your page has to be reachable by OpenAI’s search crawler, OAI-SearchBot, and it has to answer a question that makes ChatGPT go to the web at all. When we asked ChatGPT 100 blogging questions, it cited nobody on 48. Tool questions were searched every time, and platform questions 11 times in 15. Getting-started questions almost never were.
Figures read from the sources linked on 5 November 2026.
That second condition is the one most guides skip, and it decides more than anything you do on the page.
The short version
- ChatGPT chooses when to search. On our 100 questions it searched for 15 of 15 tool questions and 1 of 15 getting-started questions.
- OpenAI documents four user agents, and three matter to a blogger. OAI-SearchBot decides whether you can appear in ChatGPT search, GPTBot is for training, and ChatGPT-User fetches pages when a person asks something. The fourth, OAI-AdsBot, only visits pages submitted as ads.
- Blocking GPTBot does not take you out of ChatGPT search, and allowing it does not put you in.
- When ChatGPT did cite, about a third of the pages it cited were vendor or platform documentation. Holding a fact nobody else has is the closest a blogger gets to that position.
- You can check your own standing in about 20 minutes with a manual test, plus the
utm_source=chatgpt.comtag in your analytics. - Nothing here guarantees a citation. Blogging Titan itself did not appear once in the 244 citations ChatGPT gave in our study.
Why most blog posts never get a ChatGPT citation
Most advice on this topic assumes that ChatGPT looks something up every time someone asks it a question. It does not. OpenAI’s own launch post for ChatGPT search says that “ChatGPT will choose to search the web based on what you ask”, and that a user can also press the search icon to force it.
When it chooses not to search, it answers from what the model already knows. No source is attached, so there is no citation for anyone to win. That is why so much effort on “AI optimisation” goes nowhere for bloggers. A beautifully structured post on how to write a headline competes for a slot that, on our data, usually does not exist.
So the real job has two halves. First, work out whether your topic is one ChatGPT searches for. Second, if it is, make sure the crawler can reach you and that your page is worth quoting when it arrives.
Step 1: find out whether ChatGPT searches for your topic
On 2 September 2026 we put 100 ordinary blogging questions to ChatGPT, one per fresh conversation, signed in to a free-plan account on a UK connection. We did not tell it to search. The full method and dataset are in the AI citation league table.
The split by topic was stark:
| Question type | Questions asked | ChatGPT searched and cited |
|---|---|---|
| Tools and software | 15 | 15 (100%) |
| Platforms and hosting | 15 | 11 (73%) |
| AI and AI search | 7 | 4 (57%) |
| SEO and traffic | 20 | 11 (55%) |
| Monetisation | 15 | 7 (47%) |
| Admin and maintenance | 5 | 2 (40%) |
| Writing and ideas | 8 | 1 (12%) |
| Getting started | 15 | 1 (7%) |
We call the bottom of that table the how-to blackout. Fourteen of the 15 getting-started questions came back with no source at all. “How to start a blog” is the most fought-over query in this niche, and on that day ChatGPT gave nobody credit for it.
The name could mislead, so one qualification. The blackout tracks topic, and the word “how” has little to do with it. Forty-one of our questions began with “How”, and ChatGPT searched on 16 of them, including how to speed up a slow WordPress blog and how to get a blog to rank on Google.
The pattern looks like a system that searches when the answer might be out of date or when the person is close to buying something. Prices change, tools change, and platforms change their plans. Advice on writer’s block does not.
What to do with this
Sort your posts into the two groups before you spend an hour on any of them. A pricing page, a comparison, a “best X for Y” list or a post about a tool that changed last month is in the searched group. Evergreen beginner advice mostly is not.
That does not make beginner posts worthless. They still earn Google traffic, and Perplexity treats them differently (we cover that in our guide on how to get cited by Perplexity). It means that if ChatGPT citations are your goal this quarter, start with the posts ChatGPT is likely to look up.
Step 2: let the right OpenAI crawler in
OpenAI publishes a page describing each of its bots. It lists four user agents. One, OAI-AdsBot, is described as checking the safety of pages submitted as ads on ChatGPT, so it only matters if you advertise there. Read the page once anyway, because the other three do very different jobs and most “block AI” guides treat them as one.
| User agent | What OpenAI says it is for | What robots.txt does | Requests on our blog |
|---|---|---|---|
| OAI-SearchBot | Surfacing websites in ChatGPT search | Blocking it opts you out of ChatGPT search answers | 38 page fetches in 60 days |
| GPTBot | Crawling content that may be used to train OpenAI’s foundation models | Disallowing it says your content should not be used in training | 0 in the 35 days to 14 September, 2 in the 15 days after |
| ChatGPT-User | Visiting a page during a user’s request in ChatGPT or a custom GPT | OpenAI says robots.txt rules “may not apply” because a user started the action | 2,341 page fetches in 60 days |
The right-hand column comes from the logger we run on this site. Over 60 full days, 11 August to 9 October 2026, our server recorded 2,615 page fetches from AI retrieval bots, the crawlers that fetch pages to answer a question or to build an AI search index. ChatGPT-User made 2,341 of them, or 89.5%. That log is in our piece on what Cloudflare Pay Per Use offers a small blog. GPTBot, by contrast, made no page requests at all in the 35 days before 15 September and 2 in the 15 days after, as our Cloudflare crawler change log records. That is one small blog and should not be generalised, but it is a real log and not an estimate.
What OpenAI says about appearing in ChatGPT search
OpenAI’s publisher FAQ opens with a line every blogger should read: “Any public website can appear in ChatGPT search.” There is no submission form or listing to apply for. That answers the “how to get listed in ChatGPT search” question directly: there is nothing to apply for, and the work is making sure nothing blocks the crawler.
The same FAQ tells publishers to “make sure you aren’t blocking OAI-SearchBot”, and adds that you may need to update your robots.txt file so that it has access. The bots page says that after you change robots.txt, it can take about 24 hours for OpenAI’s systems to adjust.
If you want a page kept out of ChatGPT search, the FAQ says to use a noindex meta tag. It also warns that the crawler has to be allowed to crawl the page to read that tag. Block the page in robots.txt and add noindex, and OpenAI cannot see the noindex.
A robots.txt that keeps you eligible
The simplest robots.txt that keeps you in ChatGPT search is one that does not mention OpenAI at all. Ours is a single group for every user agent, blocking only /wp-admin/. If you want to be explicit, or you block other bots and want to carve out an exception, this is the shape:
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /
That example allows search and refuses training. OpenAI’s bots page says each setting “is independent of the others”, so you can make the training decision on its own merits. We take no position on whether you should allow GPTBot. It is a choice about your content’s use in training, and it has no documented effect on ChatGPT search.
Check the firewall as well as the file
A robots.txt file only states your wishes. A firewall, a security plugin or a CDN can refuse the bot before it ever reads the file. OpenAI publishes IP lists for each bot (for example https://openai.com/searchbot.json) so that a firewall can be told to let them through. If your host or CDN has a “block AI bots” switch, find out which user agents it covers. If your site runs through Cloudflare, read our change log on the 15 September change to its AI-crawler controls. Our own blog sits outside Cloudflare, and its ChatGPT fetches barely moved.
Step 3: be the primary source for something
Once ChatGPT searches, which pages does it pick? In our study it cited 244 sources across its 52 searched answers, a median of 4.5 per answer. Its single most cited domain was google.com. Google’s own properties (Search Central, Search Console help, the Google blog and Blogger) supplied 46 of the 244, or 18.9%.
Ask ChatGPT how to get indexed faster and it cites Google’s documentation. Ask about affiliate disclosure and it cites the FTC. Documentation, standards bodies, regulators and published research together made up about 29% of its citations (70 of 244).
The page-level read tells the same story. When we went back and examined who wrote each cited page, 33.9% of ChatGPT’s cited pages were vendor or platform documentation. Perplexity’s figure was 0.9%.
A blogger cannot become Google Search Central. What you can become is the original source for one fact in your niche. If you publish a dataset nobody else holds, such as prices you logged on fixed dates, a count you made by hand, or your own server logs, then everyone who repeats the figure is repeating you. ChatGPT’s tilt towards primary sources then works in your favour.
The tilt is not absolute. ChatGPT also cited a Fiverr gig listing and a Behance portfolio in one answer about WordPress. Original data improves the odds; it does not settle them.
Make the fact easy to lift
ChatGPT quotes sentences. A finding buried in paragraph nine, phrased as “it seems that quite a few sites might”, gives it nothing to take. Put the number, the date and the sample in one plain sentence near the top. “In 60 days, ChatGPT-User made 2,341 of the 2,615 retrieval fetches on our blog” can be quoted on its own. Our own AI Citation Grader scores this “answer-first” structure, among other things.
Step 4: put a name and a date on the page anyway
Here our data contains a surprise. ChatGPT was less fussy about authorship and dates than Perplexity.
With documentation set aside, 45.4% of ChatGPT’s sources named a human author, against 71.6% of Perplexity’s. ChatGPT was more than three times as likely to cite a page that named nobody at all (40.3% against 12.0%). We call that the nameless citation. Dates followed a similar line: 39.5% of ChatGPT’s non-documentation sources carried no publication date, against 15.9% of Perplexity’s, according to our citation age gap study.
So should you strip your byline to please ChatGPT? No. These are correlations from 951 pages, and the studies do not claim that a byline causes or prevents a citation. Part of the raw gap is vendor documentation, which rarely carries names, but the gap persists with documentation set aside, so documentation does not explain all of it. And a page with a real author and a date serves readers, Google and Perplexity better. On our corpus, pages that named a human were dated 92.7% of the time; pages with no human byline managed 47.8%.
The practical rule is to sign and date your posts because it costs nothing and helps everywhere else. Just do not expect a byline to be the thing that wins you a ChatGPT citation.
Getting your brand mentioned in ChatGPT is a different problem
A mention and a citation are not the same event. A citation is a linked source attached to a searched answer. A mention is ChatGPT naming you in the text, which can happen on a closed-book answer with no search at all.
Closed-book answers come from what the model learned in training. That is the part of “how to get your brand mentioned in ChatGPT” that no blogger can steer on a timetable. GPTBot is the crawler OpenAI describes for training, so a site that blocks it is asking not to be used for future training. Beyond that, OpenAI publishes no guidance on how a brand ends up in a model’s knowledge, and we have no data on it. Anyone who promises you a mention is guessing.
What you can do is be talked about in the places that get cited. That overlaps with ordinary link building and PR, and it is slow.
How to check whether ChatGPT cites you
You do not need a paid tracker to get a first answer. You need a list of questions and an honest record.
The manual test
- Write down 10 to 20 questions a reader would type to find your best posts. Phrase them the way a person would, in plain words.
- Open a fresh ChatGPT conversation for each one. Do not tell it to search; you want the behaviour a reader gets.
- Record three things per question: did ChatGPT search, which sources did it list, and were you among them.
- Open the sources panel. ChatGPT’s launch post describes a Sources button under the answer that opens the full list, and some sources sit behind it, out of sight in the main text.
- Rephrase each question once and run it again. One phrasing on one day is a single draw from a wide spread of possible answers.
- Repeat the whole set in a month, with the same questions, so you are comparing like with like.
Keep the sheet. A single citation means little: in our study, 533 of the 652 domains cited by either engine appeared on one question only. What matters is whether you turn up again and again across a topic.
Count the visits ChatGPT sends
OpenAI’s publisher FAQ says ChatGPT “automatically includes the UTM parameter utm_source=chatgpt.com in referral URLs”. Filter your analytics by that source and you can see which posts ChatGPT readers actually click through to.
Your server logs show the other side. A ChatGPT-User request means someone’s question led ChatGPT to fetch that page. It does not prove a citation, but a page that never gets fetched is very unlikely to be cited. Most caching plugins serve bots from cache, so the request may never reach WordPress. We had to set our cache to let bot user agents through before our logger could see them.
Use the grader for the on-page side
Our free AI Citation Grader checks one post against eight on-page signals and benchmarks it against 161 page-one pages we counted by hand. It does not ask ChatGPT anything. Treat it as a readiness check: it tells you whether a page gives an engine something to quote, and the manual test tells you whether one did.
One of the grader’s signals is schema markup. Our own schema study found that schema did not predict Google AI Overview citations, so do not treat a missing FAQ block as the reason you are not cited.
What you can control, and what you cannot
Most of the levers that decide a ChatGPT citation sit outside the blogger’s reach. It is better to know which is which before you start.
| Within your control | Outside your control |
|---|---|
| Whether OAI-SearchBot can reach the page (robots.txt, firewall, CDN) | Whether ChatGPT decides to search for a given question |
| Whether the page holds an original, dated fact | Which third-party search results ChatGPT draws on |
| Whether the answer sits in a quotable sentence near the top | Account settings, memory and model version on the reader’s side |
| Byline and date | Which other pages compete for the same few slots |
| Choosing topics in the searched group | How often the source set changes from one week to the next |
On the second row of the right-hand column: OpenAI’s launch post says ChatGPT search “leverages third-party search providers, as well as content provided directly by our partners”. OpenAI does not name the search providers in that post, so we do not either.
A worked example: our own blog
Blogging Titan does everything in the left-hand column. Our robots.txt allows every bot apart from the admin folder. ChatGPT-User fetched our pages 2,341 times in 60 days. We publish original data regularly.
And in our own 100-question study, Blogging Titan did not appear once, on either engine, across 1,243 citations. We are in ChatGPT’s fetch logs and not in its answers for those questions. That is the honest picture for most small blogs: being eligible and being chosen are separate steps, and only the first one is fully in your hands.
Checklist for one post
Run this on a single post before you change anything site-wide.
- Is the post’s topic in the searched group (tools, platforms, pricing, comparisons, fast-changing SEO)? If not, set your expectations accordingly.
- Does robots.txt allow OAI-SearchBot, either by name or by not blocking it?
- Does your firewall, security plugin or CDN let OpenAI’s published IP ranges through?
- Is the page free of a noindex tag, unless you meant it?
- Does the post hold at least one fact that originated with you, with the date and sample stated?
- Is the main answer in one plain sentence within the first screen?
- Does the post carry a byline and a visible date?
- Have you run the manual test on three phrasings and recorded the result?
Frequently asked questions
These are the questions bloggers ask us most about ChatGPT citations, answered from OpenAI’s documentation and our own data.
How long does it take for ChatGPT to cite a new post?
Nobody outside OpenAI can give you a timeline, and we would distrust anyone who does. OpenAI says robots.txt changes take about 24 hours to register for search. Being crawlable is only the entry ticket, so run the manual test monthly to see whether anything has moved.
Does blocking GPTBot remove my blog from ChatGPT?
Not from ChatGPT search. OpenAI describes GPTBot as its training crawler and says each robots.txt setting is independent. Search inclusion is controlled by OAI-SearchBot, so blocking GPTBot alone leaves you eligible to appear in search answers.
Can I submit my site to ChatGPT search?
OpenAI’s publisher FAQ describes no submission process. It says any public website can appear and tells publishers not to block OAI-SearchBot. Making the page crawlable is the whole of the listing step.
How do I rank higher in ChatGPT search?
OpenAI publishes no ranking factors for ChatGPT search, and its publisher FAQ offers no ranking advice beyond crawler access. Our data suggests picking topics ChatGPT searches for and holding original facts improves the odds. Any published list of ChatGPT “ranking factors” is observation from outside.
Does ChatGPT cite Reddit?
Not in our study. Across 100 blogging questions and 244 citations, ChatGPT cited no Reddit, Quora, Stack Overflow, LinkedIn, Medium or YouTube page. Perplexity cited that group 75 times, which is one reason the two engines need different plans.
Why does ChatGPT fetch my page but not cite it?
A ChatGPT-User fetch means your page was read while ChatGPT built an answer. It may have used another source, or used yours without listing it. Our own pages were fetched 2,341 times in 60 days, and Blogging Titan still did not appear among the 244 ChatGPT citations in our 100-question study.
Conclusion
Getting cited by ChatGPT starts with a question most guides never ask: will ChatGPT search for this at all? If the answer is yes, keep OAI-SearchBot welcome, put an original fact in a plain sentence near the top, and sign and date the page. Then test, record, and test again next month. The search decision, the source pool and the competition stay with OpenAI. What stays with you is being eligible, being quotable, and knowing where you stand.
Sources
- Overview of OpenAI Crawlers (OpenAI developer documentation)
- Publishers and Developers FAQ (OpenAI Help Center)
- Introducing ChatGPT search (OpenAI)
- OAI-SearchBot IP list (OpenAI)
- The AI Citation League Table (Blogging Titan, 2 September 2026)
- The nameless citation study (Blogging Titan, 24 September 2026)
- The citation age gap (Blogging Titan, 24 September 2026)
- What Cloudflare Pay Per Use offers a small blog (Blogging Titan)
- Cloudflare crawler change log (Blogging Titan, 30 September 2026)