Aimerce Blog

ChatGPT Dropped Its UTM Tags: Fix Your AI Traffic Data

28 October 2026

ChatGPT Dropped Its UTM Tags: Fix Your AI Traffic Data

First-Party Data 101

Quick Answer: After GPT-6 rolled out on October 7 and 8, 2026, links in ChatGPT answers stopped carrying the utm_source=chatgpt.com tag and started opening without a referrer. With neither signal, analytics tools filed those visits as direct traffic, so ChatGPT traffic appeared to drop sharply while direct traffic rose by almost the same amount. OpenAI called it a bug and began rolling out a fix on October 15. The visits didn't disappear, they were mislabeled. To fix your reporting, annotate October 8 to 15, read ChatGPT and direct traffic together for that week, and judge AI channels on orders rather than on sessions alone.

Key Takeaways

ChatGPT didn't send less traffic in mid-October. Its links stopped saying where the traffic came from.
Adobe's analysis of about 500 pages found ChatGPT visits fell by roughly 31,000 a day while direct visits rose by about 26,000 a day on the same pages.
OpenAI said it was a bug and began rolling out a fix on October 15. Clicks were back near prior levels within hours on the pages Adobe tracked.
The utm_source tag was already the backup signal, because ChatGPT hides the referrer on most visits. When the backup disappeared too, there was nothing left.
Any channel you measure with a tag someone else controls can go dark overnight. Orders in Shopify don't.

If your ChatGPT traffic chart fell off a cliff in the second week of October, you're not alone, and your store probably didn't lose anything.

What changed was the label, not the traffic. For about a week, ChatGPT stopped telling your site where its visitors came from, so analytics tools did the only thing they could and called those visits direct.

This matters more than it sounds, because a lot of AI measurement, including ours, depends on that one tag. In our AI Commerce Report, we found ChatGPT already hides the referrer on 81.8% of visits, so the utm_source tag was the main way to see ChatGPT traffic at all. When it went missing too, ChatGPT went invisible.

What happened to ChatGPT's UTM tags?

Since June 2025, ChatGPT had been adding utm_source=chatgpt.com to links in its answers. That tag became the standard way to identify ChatGPT traffic, because the referrer was often missing, especially from ChatGPT's mobile apps.

GPT-6 started rolling out to paid ChatGPT plans on October 7, 2026, and to free plans on October 8. Shortly after, links in the new answer interface stopped carrying the utm_source tag. They also started opening with settings that tell the browser not to pass a referrer. With no tag and no referrer, a ChatGPT click arrived at your site looking exactly like someone who typed your URL into the address bar.

Some links still carried the tag, such as links in extra panels, so the drop wasn't total. But for most answer links, the source simply disappeared.

How do we know the traffic didn't actually drop?

Because it showed up somewhere else. Cedric Huesler of Adobe analyzed about 500 pages that had been getting at least 5% of their visits from ChatGPT. In the days after the launch, ChatGPT-attributed visits fell by roughly 31,000 a day, while direct visits on those same pages rose by about 26,000 a day. Two out of three pages that lost ChatGPT traffic saw direct traffic go up at the same time.

That's close to a one-for-one swap. Real demand didn't vanish. It got relabeled.

Reports from other sites ranged widely, with many describing ChatGPT session drops of roughly a fifth to more than half in Google Analytics, while search traffic stayed steady.

Is it fixed?

Mostly. OpenAI described it as a bug and started rolling out a fix in stages on October 15. By mid-morning that day, Huesler reported ChatGPT clicks were back to roughly 95 to 100% of where they'd been at the same hour a week earlier.

The data from October 8 to 15 doesn't fix itself, though. Analytics tools don't go back and relabel sessions, so that week will stay undercounted in every report that sorts traffic by source.

ChatGPT traffic before, during, and after

	Before October 8	October 8 to 15	After the fix
utm_source=chatgpt.com on links	Yes, on most answer links	Missing from most answer links	Back, rolling out from October 15
Referrer passed	Often missing, especially from the apps	Blocked on answer links	Often missing, as before
How analytics labeled the visit	ChatGPT or AI assistant	Mostly direct	ChatGPT or AI assistant
Shopify orders	Unaffected	Unaffected	Unaffected

How do you fix your reports?

Annotate October 8 to 15. Mark the window in every dashboard that shows ChatGPT or AI assistant traffic, with a plain note: GPT-6 rollout, ChatGPT tags and referrer missing.
Read ChatGPT and direct together for that week. If direct rose while ChatGPT fell on the pages ChatGPT usually sends traffic to, treat most of that rise as ChatGPT.
Don't compare across the window. Week-over-week or month-over-month comparisons that include October 8 to 15 will show a ChatGPT dip that didn't happen.
Check that the tags are back. Click a few links to your site from ChatGPT answers and confirm utm_source=chatgpt.com is in the URL.
Keep counting by referrer or utm_source. Neither signal alone is reliable, so your detection rule should accept either.
Watch orders, not just sessions. Your Shopify orders were never affected. If ChatGPT-assisted revenue matters to you, track the orders those visits produce, not only the visit counts.

What this tells you about measuring AI channels

I don't think this was anyone's fault in a meaningful way. But it's a useful reminder of how fragile AI traffic measurement is right now. One tag, added by one company, was holding up most of the AI attribution in the industry. When it disappeared, every dashboard built on it went quiet at once, and nobody could tell from the dashboard alone whether demand had dropped or the label had.

That argues for measuring AI channels on things you control. Your Shopify orders are yours, and they don't depend on whether a referrer or a tag arrives. For paid ChatGPT Ads, the click ID captured when someone clicks an ad is a stronger link to the order than a source tag on an organic answer. Aimerce captures OpenAI's ad click ID and keeps it across sessions, and sends purchases to the ad platforms server-side from Shopify's order data.

The reports we've seen describe organic answer links. It hasn't been documented whether clicks on paid ChatGPT ads were affected the same way, so check your ChatGPT Ads data for that week separately.

FAQ

Why did my ChatGPT traffic drop in October 2026? After GPT-6 rolled out on October 7 and 8, ChatGPT links stopped carrying the utm_source=chatgpt.com tag and started opening without a referrer, so analytics tools labeled those visits as direct. The traffic was mislabeled, not lost.

Is ChatGPT adding UTM tags again? Yes. OpenAI called the change a bug and began rolling out a fix on October 15, 2026. Reported ChatGPT clicks returned close to prior levels that day on the pages Adobe tracked.

Can I recover the ChatGPT traffic that showed up as direct? Not exactly, because analytics tools don't relabel past sessions. Annotate October 8 to 15 and read ChatGPT and direct traffic together for that week, especially on pages ChatGPT usually sends visitors to.

Why does ChatGPT traffic show up as direct even without this bug? ChatGPT often doesn't pass a referrer, especially from its mobile apps. In Aimerce's AI Commerce Report, 81.8% of ChatGPT visits had no referrer, which is why counting by utm_source as well as referrer matters.

Were my Shopify orders affected? No. Orders were recorded in Shopify as usual. Only the label on the visit that led to them changed.

Did this affect ChatGPT Ads tracking? The reports describe links in ChatGPT answers. It hasn't been documented whether paid ChatGPT ad clicks were affected, so check your ChatGPT Ads data for October 8 to 15 separately.