Are You Getting Visits from ChatGPT? How to Check and What the Data Tells You

“Apparently, people are visiting us from ChatGPT.” Great—but are they?

Almost every business owner hears this claim nowadays. Someone mentioned it over lunch, or they read it in a newsletter. What to do about it is a simple question with an equally simple answer: check your analytics. A visit from ChatGPT is neither magic nor a new kind of metric. It is an ordinary click from another website—except that the other website is a chat.

It takes five minutes. The harder part is understanding what the table tells you. Zero visits do not mean that ChatGPT knows nothing about you, while a few dozen sessions do not necessarily mean that AI is making you money.

First, clarify what you are actually asking

When people talk about “appearing in ChatGPT,” they mix together two different questions. Analytics can answer only one of them.

AI traffic therefore proves that someone actually clicked. It does not measure your visibility in AI. No analytics platform can show you that because it happens in someone else’s conversation, which you cannot access. The article What ChatGPT Says About Your Company—and How to Find Out explores this distinction in more detail.

Where to find these visits in GA4

You can follow these steps without assistance:

1. Open Reports in the left-hand menu
2. Go to Acquisition → Traffic acquisition
3. Above the first column in the table, change the dimension to Session source (or Session source/medium)
4. Enter chatgpt in the search field above the table
5. Repeat the search for perplexity

If the instructions do not match what you see on screen, check which version of analytics they describe. Google discontinued the older Universal Analytics in 2023, and its data is no longer available. GA4 is now the only current version, and its reports and metric names differ. Plenty of instructions featuring old screens are still circulating online.

Search for parts of names rather than complete rows. Depending on how a link was opened, you may encounter chatgpt.com, perplexity.ai, or the older chat.openai.com. If you search for an exact source-and-medium value, you will miss some of the records.

Do not stop at those two names. Other assistants can also link to sources, so use the same procedure to search for gemini.google.com, copilot.microsoft.com, and claude.ai. Only the term in the search field changes.

Identifying a visit is not automatic. Links from ChatGPT include the utm_source=chatgpt.com parameter added by OpenAI, which helps analytics recognize the source. Without such a marker, many of these visits would be absorbed into direct traffic. Do not assume that other assistants apply a similar marker; some of their clicks will end up classified as direct traffic.

If you cannot see the Traffic acquisition report in the menu, it is not broken—someone has removed it from the navigation. A user with the Editor role can restore it when customizing the reports.

What to display so the numbers mean something

The total number of sessions is the least interesting figure in the table. Add Landing page + query string as a second dimension, and you will see what matters: which page people arriving from AI visited first.

Monitor:

• sessions and engaged sessions,
• landing pages,
• key events—form submissions and clicks on phone numbers or email addresses.

The landing page tells you more than the total number of visits. If people arrive at a specific guide or service page, you have a clue about the type of content generating those clicks—and what you might create more of. Treat it as a hypothesis, however, not a proven fact. GA4 shows only the page someone clicked. Analytics cannot tell you whether AI actually cited it as a source or whether the visitor reached it by another route. If people arrive at a general article with no path leading anywhere else, you have a visit but no enquiry. Internal Links: How to Connect Articles and Service Pages explains how to create that onward path.

Why visits from ChatGPT will never appear in Search Console

This is the most common dead end. Search Console measures a website’s performance in Google Search—how often your page appeared in the results and how many people clicked it. It is not a report of all website traffic.

ChatGPT and Perplexity are services outside Google. From an analytics perspective, clicks from them are ordinary referral traffic and belong in GA4. You will not find them in Search Console, and their absence is neither a missing setting nor a broken integration.

For the same reason, a claim you often hear today does not necessarily hold: “Our clicks from Google fell because people are searching for us through AI.” The two trends may be unrelated. The cause could be a rewritten title, new competition on the first page, or seasonality. Monitor them separately instead of treating one as the explanation for the other. Search Console for Business Owners explains how to interpret Search Console without marketing jargon.

Why GA4 may show zero even when ChatGPT recommends you

A zero in the table means only one thing: during the selected period, GA4 did not identify a visit with a matching source. It does not mean that AI never mentioned you.

The most common reasons why a visit is missing or never happens are:

• Nobody clicked. The person received an answer in the chat and was satisfied with it. The mention happened, but the click did not.
• The source was lost. The visit fell under (direct) / (none), which is how GA4 labels sessions whose source it cannot determine. This can happen after a redirect, when a link opens from an app, or when the visitor blocks tracking.
• Tracking is incomplete. The GA4 code is missing from some pages, or the cookie banner starts analytics only after consent, which many visitors do not provide.
• There is too little data. A website with only a few hundred visits a month simply does not have enough data to establish a trend.
• The period is too short. You are checking one week when a quarterly view would be more meaningful.

There is also a trap when looking at server logs: records generated by OpenAI bots are not visitors. The bots also serve different purposes and should not be confused:

• GPTBot downloads content for model training.
• OAI-SearchBot collects pages for search, which ChatGPT uses to add links to its answers.
• ChatGPT-User retrieves a page when a specific conversation requests it.

None of them belongs in the number of people arriving from ChatGPT. They are server requests without a browser, so they do not appear in GA4. That is the difference between logs and analytics: logs also record bots, whereas GA4 records people for whom the tracking code loaded. Combining the two produces a rather impressive piece of nonsense.

This distinction also has practical consequences beyond statistics. Blocking GPTBot in robots.txt concerns model training, whereas blocking OAI-SearchBot reduces the chance that ChatGPT will offer your website as a source in an answer. Before copying a block of rules from the internet, check which bot you are actually excluding.

How to tell whether the traffic is valuable

For a small business, it makes little sense to assess a source by volume. One visit that leads to a phone call can outweigh fifty article views. Work through these four questions in order:

1. Which pages do people visit? Are they related to what you want to sell, or are they incidental articles at the edge of your field?
2. Do visitors continue elsewhere? Do they move from the article to a service, pricing, testimonials, or contact page?
3. Do they take a measurable action? Do they submit a form, click a phone number, or open an email link?
4. Does the visit generate an enquiry? You can verify that only through email, your CRM, or a phone call—analytics cannot confirm it for you.

The third point often causes problems. A generic form_submit event in GA4 is not enough if the website contains three different forms—for example, an enquiry form, a newsletter sign-up, and an appointment request. You will not know which one was submitted. The solution is a separate key event for enquiries, such as lead_form_submit. Forms embedded through an iframe or hosted on another domain also usually require additional configuration through Google Tag Manager; otherwise, the submission may not reach analytics at all.

This is precisely where people tend to get stuck—and where we can take over. An Initial analysis reviews your tracking and content and determines whether the problem is that AI does not mention your company or that visitors arrive but have nothing useful to follow.

A saved exploration instead of searching every month

A one-off check after reading this article will not reveal a trend. Create an exploration in GA4 that you can open with the same settings each time.

Under Explore, create a free-form exploration and configure:

• rows: Session source and Landing page + query string
• values: Sessions, Engaged sessions, Key events
• filter: source contains chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai, and optionally chat.openai.com

Compare the last 90 days with the previous 90 days. For a small business, one month often contains so little data that a single additional visitor looks like a trend.

Then record five lines every month—even in a notebook:

People often omit the final line and then struggle to explain fluctuations. A simple note such as “Service page rewritten on 18 June” can make the chart understandable three months later. How to Tell Whether Content Pays Off explains in more detail what to monitor and what to ignore.

How much this tracking costs

You do not need to buy anything for a basic check. GA4 is free and sufficient for this purpose—provided it tracks form submissions and clicks on phone numbers, not just page views.

The cost lies elsewhere: in the initial setup and, above all, in verifying that events are actually being recorded. It is worth doing. Six months of decisions based on incorrect data will cost a small business more than one thorough review of GA4, the cookie banner, and Tag Manager.

Paid “AI visibility” tools solve a different problem. They repeatedly submit queries and monitor whether you appear in the answers. They can be useful, but they do not measure a single visit or enquiry. They complement GA4; they do not replace it.

What to do based on what you found

You can see visits from AI. Find out which pages attract them and review the path from those pages to your services and contact details. Do not automatically commission five more general articles on the same topic. Instead, add whatever the successful pages are missing: prices, the process, and specific answers.

You cannot see any visits. Follow this order so you do not address content before checking the tracking:

1. verify that GA4 is working and recording key events,
2. extend the period to 90 days,
3. check other AI sources, not just ChatGPT,
4. find out whether AI mentions you in response to relevant questions,
5. only then address content gaps.

That order is not a formality. Without it, you can easily end up writing more articles to solve a problem that actually lies in the cookie banner. And if AI genuinely does not mention you, that is a content problem. How ChatGPT Decides Whom to Recommend explains what guides its choices and what it prioritizes in answers.