ChatGPT dropped its tracking tags, so your AI traffic numbers are probably wrong

ChatGPT stopped adding tracking tags and referrer data to the links it shows in answers, starting around 8 October 2026 after the GPT-6 rollout. This meant visits that should have shown up as "ChatGPT" traffic in tools like Google Analytics 4 were instead logged as "Direct" traffic, making it look like AI referrals had collapsed when they probably hadn't. OpenAI has since shipped a partial fix, but the week of broken data is already baked into everyone's reports.

If your dashboards showed ChatGPT traffic falling off a cliff this month, this is almost certainly why.

What changed?

ChatGPT links used to carry two clues that told your analytics where a visitor came from: a utm_source=chatgpt.com tag on the URL, and a referrer header passed by the browser. Adobe's Cedric Huesler noticed OpenAI appears to have removed UTM tracking parameters from links in ChatGPT, potentially causing visits from the AI platform to be incorrectly classified as direct traffic, with the change spotted following the release of GPT-6 in October 2026.

With both clues missing, analytics platforms had nothing left to work with. As one analysis put it, Google Analytics has nothing to go on, so it reports the visit as (direct) / (none), as if the person had typed your address from memory.

Huesler put numbers on it. He examined approximately 500 pages where ChatGPT previously accounted for at least 5% of visits, comparing 9 to 13 October with the preceding week, and found ChatGPT traffic down approximately 31,000 visits per day and direct traffic up approximately 26,000 visits per day on the same pages. Crucially, two-thirds of pages losing ChatGPT referrals experienced an increase in direct traffic, which is the clearest sign that visits moved columns rather than disappearing.

The behavioural evidence backs this up too. On mobile, the share of Direct visitors who saw the consent banner, his sign of a first visit, went from 36% to 44%, moving toward the 55% he sees among ChatGPT visitors, according to Search Engine Journal.

Not everyone agrees it's purely a labelling issue, though. One analysis found something odder: Sébastien Billard recorded a drop since 9 October not only in human traffic from ChatGPT but also in visits from the ChatGPT-User bot, while OAI-SearchBot activity has been rising since 7 October, suggesting it's possible that how often ChatGPT fetches pages during a conversation has changed too, per Delante. That part still needs watching.

Has OpenAI fixed it?

Partly, and quickly. According to reporting pulled together by Ignite Visibility, OpenAI reportedly confirmed the tracking issue on 14 October, and a fix began rolling out on 15 October, with one analysis finding tagged ChatGPT clicks had recovered to roughly 95% to 100% of the same hour the previous week by 9 a.m. PT that day.

But the fix isn't complete. Search Engine Journal notes Huesler says ChatGPT started opening links in the answer text without the utm_source=chatgpt.com tag or a referrer, so visits from these links can appear as Direct in GA4, and there are other analyses suggesting the same change in tracking. A separate write-up points out that the UTM tag is back, but the referrer is not: on 15 October, Huesler reported that OpenAI had told him the missing tag was a known bug, and that tagged ChatGPT clicks were already back to roughly 95 to 100% of the previous week's level, though the fix may not have reached every account yet, and it brings back the tag, not the referrer.

In plain terms: the tag fix helps, but without the referrer coming back fully, some ChatGPT visits will keep landing in the Direct bucket for a while yet.

Why does this matter for your business?

If you run paid media, content or GEO strategy and you use ChatGPT referral volume as a signal of whether your brand is being surfaced and clicked, this week's numbers are unreliable. A real drop in AI citations would look identical in your dashboard to a tracking bug, unless you know to check for the second explanation.

This matters more than it might seem, because AI referral traffic converts unusually well. Separate research cited by one analysis found AI search visitors convert 4.4x better than traditional organic visitors according to Semrush, with Ahrefs finding AI visitors produced 12.1% of signups from just 0.5% of traffic. Misreading a tracking glitch as a real decline could push teams to pull budget or deprioritise GEO work based on a false signal, right when AI referrals are becoming one of the highest-value channels you have.

It's also a reminder of how fragile AI referral measurement is generally. Unlike Google Search Console, which gives you query-level data for organic search, there's no equivalent official reporting layer for ChatGPT citations and clicks. You're relying on third-party UTM and referrer conventions that a single OpenAI release can quietly break.

What should you do now?

First, don't panic-read your October ChatGPT numbers in isolation. Mark roughly 8 to 10 October on your ChatGPT referral trend so no one reads a sudden drop or a Direct spike as a performance change, something Huesler and Profound both recommend.

Second, look at ChatGPT and Direct traffic together rather than in isolation. Delante suggests you compare both channels before 7 October and after 9 October at the landing page level, paying particular attention to the homepage and the subpages that previously got traffic from ChatGPT, as well as to new users in Direct.

Third, check total traffic and conversions rather than channel labels alone. If overall numbers on a page held steady while only the "ChatGPT" line dropped, that's a strong sign of misattribution rather than lost visibility.

Finally, treat this as a prompt to audit how exposed your business actually is to AI referral channels and citations in the first place, since channel-level reporting is clearly not something you can take for granted. Running a free check with Sited's AI visibility audit gives you a view of how often your brand is actually surfacing in AI answers, independent of whatever your analytics platform happens to be labelling that week.

Frequently asked questions

Did ChatGPT traffic actually fall, or was it just a reporting glitch?

The evidence points mostly to a reporting glitch. Two-thirds of pages losing ChatGPT referrals experienced an increase in direct traffic, and the new Direct visitors behaved like typical ChatGPT visitors, suggesting most of the "lost" traffic simply got mislabelled rather than disappearing.

When did this start and is it fixed now?

It started around 7 to 9 October 2026, following the GPT-6 rollout. OpenAI confirmed the issue on 14 October and began rolling out a fix the next day, though the referrer data has not fully returned, so some misattribution may continue.

What should I check in my own analytics?

Compare your ChatGPT and Direct traffic channels for the same landing pages before and after 8 October. If ChatGPT fell and Direct rose by a similar amount on the same pages, you're almost certainly looking at the tracking bug rather than a genuine drop in AI visibility.

Does this affect other AI platforms like Gemini or Perplexity too?

The reporting so far is specific to ChatGPT and tied to the GPT-6 release, with no equivalent issue reported for Gemini, Perplexity or Copilot at this point. That said, it's a useful reminder that any AI platform could change link behaviour without notice, so it's worth monitoring referral patterns across all of them regularly.

Why does this matter if my ChatGPT traffic volume is small?

Even small volumes of AI referral traffic tend to convert well, with research showing AI search visitors convert 4.4x better than traditional organic visitors. Losing sight of that channel, even temporarily, can lead to under-investing in a high-value audience.