How to Check if ChatGPT Is Citing Your Site, Without Paying for a Tool

Introduction

There is a specific frustration in reading about AI search visibility.

Every guide tells you it matters. Most of them then tell you to enter your domain into their free checker, hand over your email, and start a trial.

The manual version of this is not difficult. It takes about twenty minutes to set up and ten minutes a month to run, and it teaches you something the dashboard never will, which is what "cited" actually looks like for your topic.

This guide covers the three free methods, what each one can and cannot tell you, and the point at which paying for a tool starts to make sense. Some of what follows is less encouraging than the tool marketing suggests, which is the point.

The Three Things You Are Actually Measuring

People collapse these into one question. They are separate, and they fail separately.

The first is whether OpenAI can reach your site at all. This is a crawler and robots.txt question, and if the answer is no, nothing else applies.

The second is whether ChatGPT actually uses your content. This is the citation question. It happens inside the answer, and no analytics tool will show it to you.

The third is whether any of it sends you visitors. This is the referral question, and it is the only one of the three that appears in your analytics.

You can pass the first, fail the second, and never reach the third. Checking only one gives you a misleading picture, which is why the methods below cover all three.

Method One: Ask ChatGPT Directly

The simplest method, and the one that shows you what an actual user sees.

Build a fixed prompt list

The instinct is to ask "what is WhatNWhy" or to search your own brand. That tells you very little. A brand query proves ChatGPT knows your name, not that it recommends you when someone has a problem.

Write ten to fifteen prompts phrased the way a reader would ask, before they know your brand exists. For a digital marketing site, that looks like:

• How do I fix a page that is indexed but not ranking
• What is a realistic PPC budget for a small UK business
• Why are my Google Ads conversions not recording
• How much should I spend daily on Facebook ads
• What is Meta Advantage Plus Creative

Write them down and keep the list fixed. The value comes from asking the same questions repeatedly over months, not from asking clever ones once.

Run them properly

• Use a logged out session or a temporary chat, so past conversations do not influence the answer
• Turn on web search if your account offers it as a toggle, since citation behaviour differs between the model answering from training and the model searching live
• Ask each prompt in a fresh chat, not as follow ups in one conversation
• Record the date, the exact prompt, the sources named, and whether you appeared

Record it in a spreadsheet

Six columns is enough:

That spreadsheet is the entire product that AI visibility tools sell you. Theirs is faster and prettier. Yours is free and you understand exactly how it was built.

What this method cannot tell you

Responses are not deterministic. Ask the same question twice and you can get different sources. This is the difference between an AI assistant and a search engine, and it means a single check proves nothing either way.

Run each prompt two or three times before recording a result, and treat the pattern across months as the signal rather than any individual answer.

It also does not scale. Fifteen prompts checked monthly is manageable. Two hundred prompts checked weekly is not, and that is precisely where a paid tool earns its money.

Method Two: Check Your Analytics for Referral Traffic

This is where you find out whether citations turn into anything.

Where to look in GA4

When ChatGPT links to a page, it appends a tracking parameter to the URL. Analytics then records the visit as a referral from chatgpt.com rather than as direct traffic.

To find it:

1. Open Reports, then Acquisition, then Traffic acquisition
2. Change the primary dimension to Session source / medium
3. Type chatgpt into the search filter

Anything present will appear as chatgpt.com / referral.

Make it a permanent channel

Filtering by hand every time is tedious, and AI traffic gets buried among ordinary referrals. A custom channel group fixes that.

In GA4, go to Admin, then Data display, then Channel groups. Copy the default group, add a new channel, and set the Session source condition to match a pattern covering the main AI platforms.

Two things are worth knowing before you build it. Filter on source rather than medium, because ChatGPT's referral URLs include a source parameter but often no medium or campaign, so a filter built on medium will miss sessions. And order matters, because GA4 assigns a session to the first channel it matches, so your AI channel needs to sit above Referral in the list.

Custom channel groups apply retroactively, so you will see historical data immediately.

The number will be lower than you expect, for a reason

Two things suppress it, and both are worth understanding before you conclude that AI search is not working.

Mobile app links drop the referrer. When someone taps a link inside the ChatGPT mobile app, it opens in an embedded browser that does not pass the referring page. Copying and pasting a link does the same. Those sessions land in Direct, and you cannot separate them from anyone else who typed your URL.

Citations without clicks also do not appear anywhere. If ChatGPT uses your content to build an answer and the reader never clicks, that visibility is real and completely invisible to analytics. This is the same problem publishers report with AI Overviews.

So treat the referral number as a floor, not a measurement. Your actual influence is higher than the number and you cannot know by how much.

Method Three: Check Whether OpenAI Can Reach You

This is the check people skip, and it is the one that most often explains a total absence.

The three crawlers, and why they are different

OpenAI operates three separate user agents, each with a different job. Their documentation states that the settings are independent, so you can allow one and block another.

This distinction matters because a lot of sites blocked OpenAI in 2023 and 2024 without realising they were blocking three different things at once. If visibility in ChatGPT is your goal, blocking OAI-SearchBot works directly against it.

Check your robots.txt

Open yoursite.com/robots.txt and look for any rule blocking those three names. If you find one and you want AI visibility, remove it.

OpenAI notes that after a robots.txt change, it can take around twenty four hours for search eligibility to update.

Check your server logs

Your logs tell you which pages OpenAI has actually fetched, which is more reliable than any inference from the outside.

Search your access logs for:

OAI-SearchBot
GPTBot
ChatGPT-User

Most hosting panels expose raw access logs. On Hostinger, cPanel, or similar, look for a Raw Access Logs or Log Manager section.

What you are looking for:

• Are any of the three appearing at all
• Which URLs are they fetching
• Are they receiving 200 responses, or 403 and 429 errors

That last one is worth checking carefully. A firewall or rate limiting rule can return 429 Too Many Requests to a crawler even when robots.txt permits it, which produces the same outcome as a block while looking fine in robots.txt.

Verify the crawler is genuine

Anyone can put GPTBot in a user agent string, and scrapers do. OpenAI publishes the IP ranges each crawler uses, so you can confirm whether a request came from OpenAI or from something pretending to be it.

The published files are linked in the resources section below.

For most small sites this is optional. It becomes worth doing if you see unusually heavy crawling and want to know whether it is really OpenAI before you block it.

Reading the Results Together

Each method answers a different question, and the combination tells you where the problem is.

The second row is where most sites sit, and it is the honest starting point for a new site. OpenAI can read you, and there is no particular reason for it to choose you over a source it already knows.

When a Paid Tool Is Actually Worth It

The free methods have real ceilings, and pretending otherwise would be as unhelpful as the tool marketing.

When a tool makes sense

• You are tracking more than roughly twenty prompts
• You need to check across ChatGPT, Perplexity, Gemini, and Copilot rather than one platform
• You need historical trend data you did not think to start collecting months ago
• You are reporting to a client or a manager who wants a chart
• Someone else's time is more expensive than the subscription

When it does not

• You have one site and a handful of topics
• You are checking monthly rather than daily
• Your site is new enough that the answer is going to be no for a while regardless

That last point deserves saying plainly. If your site is a few months old with limited authority, a tool will accurately and expensively tell you that ChatGPT is not citing you. The manual method tells you the same thing for nothing.

What Actually Increases Your Chances

This is not the focus of this guide, but the diagnosis is useless without it.

Be reachable. Allow OAI-SearchBot, and confirm your pages return 200 to it.

Answer the question near the top. AI systems extract passages, and a page that buries its answer under six paragraphs of preamble is harder to lift a clean quote from.

Use question shaped headings, because they map onto how people phrase prompts.

Cite your own sources. Content with named, checkable sources gives a system more reason to treat it as reliable.

Cover the topic properly. Breadth across related questions appears to matter more than depth on one page, which is an argument for topic clusters rather than isolated articles.

And be indexed in the first place. This one is foundational and easy to overlook. Our guide on why Google may not be indexing your page covers the diagnosis, and our guide on optimising posts for AI Overviews covers the structural side in more detail.

None of these guarantee a citation. Anyone offering a guaranteed method for appearing in AI answers is describing something Google and OpenAI have not documented and cannot be verified.

Monthly Check Routine

Ten minutes, once a month.

• Run your fixed prompt list, two or three attempts each
• Record which sources appeared and whether you were among them
• Note which competitors appeared, and which of their pages
• Open GA4 and check referral sessions from chatgpt.com
• Note which landing pages received them
• Scan server logs for the three OpenAI user agents
• Confirm they are getting 200 responses, not 403 or 429
• Check robots.txt has not changed, particularly after a plugin update

That last item catches more problems than it should. Security and SEO plugins update their default rules, and blocked AI crawlers are a common side effect.

Common Mistakes

• Checking brand queries only. ChatGPT knowing your name is not the same as it recommending you when someone has a problem.
• Checking once and drawing a conclusion. Responses vary, so one answer is an anecdote.
• Treating referral traffic as the whole picture. Mobile app clicks and uncited reads are both invisible, so the number is a floor.
• Blocking all OpenAI crawlers and then wondering about visibility. The three crawlers do different jobs, and blocking OAI-SearchBot removes you from ChatGPT search answers.
• Ignoring 429 errors. Rate limiting produces the same result as a block, while robots.txt looks perfectly fine.
• Paying for a tool before running the manual check. You will not know what the dashboard is measuring until you have done it by hand once.
• Optimising for AI before being indexed. If Google has not indexed a page, it is not in the retrieval pool that most AI search features draw from.

Conclusion

The tools are not a scam. They automate a real task and they do it well at scale.

They are also not necessary for most people reading this. If you run one site and care about a handful of topics, a spreadsheet, ten minutes a month, and your server logs give you the same answer.

Start with the crawler check, because it is fast and it catches the problem that makes everything else pointless. Then build your prompt list and run it. Then look at your analytics, knowing that the number understates reality.

If the answer comes back as no, that is normal rather than a failure. Citation follows authority, and authority follows time.