ChatGPT Appears To Drop UTM Tracking, Making AI Referral Traffic Harder To Measure

OpenAI appears to have removed UTM tracking from links in ChatGPT. If this is true, visits from ChatGPT may be counted as direct traffic instead of being recorded as visits referred by ChatGPT.

The change appears to have happened after GPT-6 launched in October 2026. It has raised questions among marketers and SEO experts about whether reports accurately show how much website traffic comes from AI tools.

Anu Adegbola of Search Engine Land spotted the pattern on October 15. The findings suggest that ChatGPT's popularity may not match the amount of website traffic credited to it in reports. So far, there is no clear sign that websites are actually losing large numbers of visitors.

What Changed In ChatGPT Links

Links from ChatGPT appear to have stopped including the utm_source=chatgpt.com tag that OpenAI added in June. UTM tags are extra information added to a web address. Analytics tools read these tags to identify where a visitor came from.

Missing UTM tags may not be the only reason ChatGPT visits are harder to track. Browsers can also send information about the website or service that referred a visitor. Cedric Huesler of Adobe reported that ChatGPT links are now opening in a way that hides this referral information.

Both changes together create the problem. If either the UTM tag or the referral information is missing, analytics tools may count a visit as direct traffic instead of crediting it to ChatGPT.

What Adobe’s Numbers Show

Huesler studied about 500 pages where ChatGPT had previously accounted for at least 5% of visits. He compared the data from October 9 to 13 with the data from the week before. Here is what he found:

• ChatGPT traffic: About 31,000 fewer visits per day.
• Direct traffic: About 26,000 more visits per day across the same pages.
• Search traffic: Down 7%, matching the wider trend in search traffic.
• Page patterns: Two thirds of the pages that lost ChatGPT referrals saw their direct traffic increase.

The numbers do not match exactly. ChatGPT traffic fell by about 31,000 visits per day, while direct traffic rose by around 26,000. That leaves a gap of about 5,000 visits. The analysis does not fully explain this difference, partly because it covers only a specific group of pages.

There is also another clue. Huesler found that the newly counted direct visitors behaved in ways similar to visitors previously identified as coming from ChatGPT. This suggests that many of these visitors may still be coming from ChatGPT, even though the analytics tools now place them in a different category.

How Far The Adobe Findings Stretch

The numbers do not match exactly. ChatGPT lost about 31,000 visits per day, while direct traffic gained around 26,000. This leaves a difference of about 5,000 visits. The analysis does not fully explain the gap because it covers only a specific group of pages.

There is another piece of evidence that supports this explanation. Huesler found that the newly counted direct visitors behaved similarly to visitors previously identified as coming from ChatGPT. In other words, many of these new direct visitors may be the same kind of visitors as before.

Why This Looks Like An Attribution Problem

The most likely explanation for this unusual pattern is a tracking change, not a drop in demand. Search traffic fell by only 7%, matching the wider trend. That suggests the sharp drop in ChatGPT referrals is not part of the general search decline. At the same time, direct traffic rose on pages where ChatGPT referrals had fallen. A change in tracking could explain both patterns because it would affect how visits from ChatGPT are counted and how much traffic is reported as direct.

This is the main idea behind Search Engine Land's analysis, although the publication has been careful not to treat it as proven. OpenAI has not confirmed that it made this change. The analysis also does not prove that all missing ChatGPT referrals were counted as direct traffic. Some visits may have been assigned to other sources. Several questions remain: does this affect every link from ChatGPT, will the change last, and will OpenAI comment on it? Until OpenAI provides answers, the change should be treated as unconfirmed.

What It Does To Your Dashboards

For people tracking how visitors discover websites through AI, one risk is that reports may give the wrong impression. For example, ChatGPT referrals might fall while direct traffic rises at the same time. This could mean that the tracking system has changed, even if the dashboard does not warn users that anything is different.

This can lead to wrong conclusions. Teams might think ChatGPT has stopped sending visitors or that a piece of content has become less popular. In reality, visitors may still be arriving but being counted under a different category.

Direct Traffic Is A Catch-All

Analytics tools use direct traffic for visits when they cannot identify where the visitors came from. This includes people who type a web address or use a bookmark, as well as people who click links in apps or documents that do not share referral information. If ChatGPT visits are added to this group, direct traffic may look stronger than it really is, while ChatGPT traffic may look weaker.

Talking To Clients And Stakeholders

Anyone reporting on traffic from AI tools may soon face a simple question: why have visits from ChatGPT fallen? The most honest answer for now is that some of the visits may have been counted incorrectly.

It is also helpful to show the evidence. A report that includes the drop in ChatGPT traffic, the rise in direct traffic on the same pages, the wider search trend for comparison, and what is still unknown will be more convincing than a report that leaves people surprised later.

What Marketers Can Do While This Plays Out

• Read the trends together. Look at referral traffic, direct traffic, and landing-page performance together before deciding whether ChatGPT is performing well.
• Check your own pages. You can try a version of Huesler's test. Find pages where ChatGPT used to account for a large share of visits and check whether direct traffic rose around the same time that ChatGPT traffic fell. If it did, your results may look similar to Adobe's findings.
• Hold off on big decisions. If visits from ChatGPT have dropped sharply, pause and investigate before removing content or changing your plans for reaching people through AI tools. Making those changes based on this data alone would be too soon.
• Watch for an update. If OpenAI changes this behavior or returns to its earlier approach, note the date. Your reports may then show ChatGPT traffic rising again.

The Wider Measurement Picture

AI assistants are becoming another way for people to find and visit websites. It is becoming more important to measure these visits accurately. Changes in how AI platforms handle links to other websites can make this harder, even when the total number of visitors stays fairly steady.

For SEO teams, referral data from AI platforms helps show whether being visible in AI tools leads to website visits. That data is now becoming less reliable, making it harder to show the return on investment (ROI) from work aimed at helping people discover content through AI.

The difference between the reported numbers is worth noting. Search Engine Land also reported this week that OpenAI is adding measurement partners and has redesigned the Conversions page for ChatGPT Ads. While tools for measuring paid ads are expanding, tracking visits from links in ChatGPT is becoming more difficult.

Search Engine Roundtable has also reported on the missing UTM tags, and Search Engine Land linked to that article for readers who want to learn more.

Frequently Asked Questions

Q1: What appears to have changed with ChatGPT links?
Links from ChatGPT appear to have stopped including the utm_source=chatgpt.com tag that OpenAI added in June. Adobe's Cedric Huesler also reported that the links are now opened in a way that hides referral information. Because of this, analytics tools may count visits as direct traffic.

Q2: Why would ChatGPT visits show up as direct traffic?
If both the UTM tag and referral information are missing, analytics tools may count a visit as direct instead of crediting it to ChatGPT. Direct traffic is the category used for visits when the tools cannot identify their source.

Q3: What did Adobe’s analysis find?
Cedric Huesler studied about 500 pages where ChatGPT had previously accounted for at least 5% of visits. He compared October 9 to 13 with the week before. ChatGPT traffic fell by about 31,000 visits per day, while direct traffic rose by about 26,000. Search traffic fell by 7%, matching the wider trend.

Q4: Does this mean ChatGPT traffic actually dropped?
Not necessarily. The findings suggest that visits may be counted under the wrong source rather than that fewer people are visiting. Two thirds of the pages that lost ChatGPT referrals saw direct traffic rise. However, the analysis does not prove that every missing referral has the same cause.

Q5: Has OpenAI confirmed the change?
No. OpenAI has not confirmed whether it removed the tracking information or whether the change was intentional. Search Engine Land describes the change as something that appears to have happened. The issue was noticed after GPT-6 launched in October 2026.

Q6: How can I tell if my site is affected?
Check pages where ChatGPT used to account for a meaningful share of visits. See whether direct traffic rose around the same time that ChatGPT traffic fell. Review referral traffic, direct traffic, and landing-page performance together before deciding what the numbers mean.

Q7: Should marketers change their AI strategy because of this?
Not yet. A sudden drop in ChatGPT referrals is a reason to investigate until the tracking behavior is better understood. Cutting back on work to help people discover content through AI based on these numbers alone would be premature.

Q8: What is a UTM parameter?
It is a tag added to a web address that analytics tools use to identify where a visit came from. OpenAI added utm_source=chatgpt.com in June to mark visits from ChatGPT.

Q9: Why does attribution matter more now?
AI assistants are becoming another way for people to discover websites. Accurate tracking helps businesses measure their impact. Changes in how platforms handle links to other websites can make this harder, even if the number of visitors stays about the same.

Q10: Does the ChatGPT drop match the direct increase exactly?
No. Adobe's figures show about 31,000 fewer ChatGPT visits per day and 26,000 more direct visits, leaving a difference of around 5,000. The analysis does not explain this gap and covers only a specific group of pages.