Between 8 and 15 October 2026, ChatGPT stopped adding its utm_source=chatgpt.com tag to many links after the GPT-6 rollout, so visits it sent could show up as Direct traffic in Google Analytics 4. Search Engine Journal reported on 15 October that one estimate put the drop in ChatGPT-attributed sessions at roughly 45%, while Adobe's data showed Direct traffic rising on the same pages. A fix reportedly began rolling out on 15 October. If you are a UK owner who checks GA4 to see whether AI search is worth the effort, last week's numbers are probably wrong. How you handle that week decides whether you make a bad call about your marketing.
What happened to ChatGPT traffic in October 2026?
According to Search Engine Journal's report, Graphite's Ethan Smith saw ChatGPT-attributed GA4 sessions fall by 22 to 57% across the sites it tracks from 9 October, and estimated a drop of about 45% overall. LLM Pulse found ChatGPT sessions about 46% lower on 9 to 11 October than on the same days in the previous two weeks, while total site traffic stayed roughly constant. LLM Pulse did not see Direct rise, and said its aggregated data could not separate lost visits from relabelled ones.
The clearest signal came from Adobe. Cedric Huesler compared 9 to 13 October with the previous week across about 500 pages where ChatGPT drove at least 5% of visits. ChatGPT visits fell by about 31,000 a day and Direct visits rose by about 26,000 a day. Most of those visitors had not vanished. They had lost their label.
Why did the visits show up as Direct?
Search Engine Journal reported that links in ChatGPT's Sources panel and shopping links kept their tags. The inline links in the answer text were the ones affected. Profound found that the share of ChatGPT answers with at least one tagged source link fell from about 98% on 8 October to 66% by 14 October. Over the same days, GPT-6's "Intelligent UI" formats, such as charts and comparisons, grew from about 2% to about 35% of the answers Profound collected.
Phone traffic was hit hardest. Smith found that 95% of mobile ChatGPT visits to Graphite's sites relied on the tag alone, because the app rarely passes a referrer. If most of your enquiries come from phones, this is the part of your data most likely to be skewed.
Is it fixed now?
Relevant Audience's timeline of Huesler's posts, published 16 October, puts the break at about 2pm Pacific on 8 October and the fix starting about 2am Pacific on 15 October. By 9am Pacific that day, tagged ChatGPT clicks were back to roughly 95 to 100% of the same hour a week earlier. The same piece notes OpenAI's release notes for 1 to 15 October do not mention the bug or the fix, and that all of these times come from Huesler rather than OpenAI.
In UK time (British Summer Time), that works out at roughly 10pm on 8 October to the late afternoon of 15 October. If your GA4 property is set to the UK time zone, treat 8 to 15 October inclusive as the affected window, which matches the window Huesler now uses.
Who in the UK should care about this?
The risk is not the lost data. The risk is acting on it. An owner who sees ChatGPT traffic halve in a week might cut content work aimed at AI answers, cancel an AI visibility tool or tell an agency the approach has failed, without noticing that Direct traffic rose in the same week. If you are still deciding how much AI search matters compared with Google, the plain guide to AEO, SEO, GEO and LLMO sets out where each one fits.
How do I check whether my own GA4 data was affected?
In GA4, open Reports, then Acquisition, then Traffic acquisition, and switch the dimension to Session source. Set the comparison dates and find chatgpt.com. Then add a secondary dimension of Landing page and repeat the check for Direct. Include your home page: Peec AI's Tomek Rudzki found that 54% of the untagged page URLs shown in GPT-6 answers pointed to home pages, according to Search Engine Journal.
Use this if/then checklist:
- If ChatGPT fell and Direct rose on the same pages: relabelling. Annotate the week and carry on.
- If ChatGPT fell and Direct stayed flat: possible real loss. Check whether ChatGPT still mentions you for your main questions.
- If ChatGPT sessions are back to late September levels from 16 October: the fix reached your traffic. Treat 8 to 15 October as a gap.
- If ChatGPT is still low after 16 October: the rollout may be uneven. Recheck weekly and do not change budgets yet.
- If ChatGPT is under 1% of your sessions: note it and move on. The week will not change any decision.
What does this look like with real numbers?
Here is the arithmetic, using made-up practice figures and Graphite's estimated 45% drop as reported by Search Engine Journal:
Total enquiries for the practice barely moved. What moved was where the credit went. If the owner pays £250 a month for content aimed at AI answers and judges it on that week, the work looks like it lost nearly half its value. Cancelling it on that basis would be a decision made on a reporting bug.
Should I set up a separate AI channel in GA4?
Start by opening Traffic acquisition with the default channel group and looking for the AI Assistant line. If it is there and picks up chatgpt.com, you already have a dedicated view. If you want more control, for example to add assistants the default rules miss, Google's GA4 help page on custom channel groups explains how to create one in Admin, under Data display, then Channel groups, by copying the default group and adding channels. You need the Editor role or higher on the property.
Order matters. Google's help page says traffic goes to the first channel whose rules it matches, so place any AI channel above Referral. Set the rule to session source matching chatgpt.com and any other assistants you see in your own source report. Search Engine Journal also reported that Gemini began adding UTM parameters to its links in late September, so check whether Gemini now appears as a source in your data. If you want help deciding which AI answers to target in the first place, see the AI visibility service overview.
What are the trade-offs of reading AI traffic at all?
Search Engine Journal pointed out that Graphite, LLM Pulse, Profound and Peec AI all sell SEO or AI search tools, and that LLM Pulse and Profound did not publish sample sizes. Their data broadly agrees on the direction, but none of it is your data. Huesler himself described his figures as early and sampled.
For a small business, the more reliable measures sit closer to the till. Ask new enquirers how they found you, add an "AI assistant" option to your contact form's source question, and track calls and form fills as conversions. If you already test page changes, the guide to A/B testing for small businesses explains why you need a stable baseline before reading any result, and a broken tracking week is exactly what spoils one.
What should I do this week?
- Monday: Add a GA4 annotation for 8 to 15 October noting the ChatGPT tagging bug.
- Tuesday: Compare 1 to 7 October with 9 to 15 October for chatgpt.com and Direct, by landing page.
- Wednesday: Check chatgpt.com sessions from 16 October against a late September weekday.
- Thursday: Confirm the AI Assistant channel is picking up ChatGPT, or create a custom channel group with an AI channel above Referral.
- Friday: Add "AI assistant such as ChatGPT" to the "How did you hear about us?" question on your forms.
- Next month: Review AI-sourced enquiries, not sessions, before changing any spend.
If you run ChatGPT Ads with your own UTM tags, check that paid clicks held steady through the bug window as well.
Irvale Studio builds measurement and lead systems for UK small businesses, so every enquiry, from Google, ChatGPT or a phone call, is tracked to a source you can trust before you spend another pound.
ChatGPT Tracking Bug Hid AI Traffic in GA4 — FAQ
Did ChatGPT send fewer visitors to websites in October 2026?
The evidence so far suggests most visitors still arrived but were mislabelled. Search Engine Journal reported on 15 October 2026 that Adobe's Cedric Huesler compared about 500 pages and found ChatGPT visits fell by about 31,000 a day while Direct visits on the same pages rose by about 26,000 a day. Huesler called this early, sampled data covering four days. The gap between the two figures means some real loss cannot be ruled out. For a small business, the practical reading is that a ChatGPT dip between 8 and 15 October is a tracking problem until your own data shows otherwise.
Is the ChatGPT tracking bug fixed?
Partly, according to one secondhand account. Search Engine Journal reported Huesler as saying OpenAI confirmed the bug to him on 14 October and that a fix began rolling out early on 15 October. By 9am Pacific time, tagged ChatGPT clicks were back to roughly 95 to 100% of the same hour a week earlier. OpenAI has made no public statement, and SEO consultant Glenn Gabe said he was not yet seeing the fix on paid or free accounts. Check your own GA4 source report for chatgpt.com from 16 October onwards before assuming it is back.
Will GA4 recover the ChatGPT visits that were counted as Direct?
No. GA4 assigns a source when the session is recorded, so sessions that arrived without the tag or a referrer stay as Direct. Restoring the tag only helps future reports. Changing your channel groups will not help either, because the missing piece is the tag itself, not how GA4 groups it. The sensible response is to annotate the dates, treat that week as unreliable for AI traffic, and compare it with clean weeks either side rather than trying to rebuild the numbers by hand.
Why did ChatGPT visits show up as Direct traffic?
Two signals went missing at once. Search Engine Journal explained that ChatGPT began opening inline answer links without the utm_source=chatgpt.com tag and without a referrer, the browser information that tells a site where a visitor came from. With neither signal, GA4 has nothing to tie the visit to ChatGPT and files it under Direct. Graphite's Ethan Smith found 95% of mobile ChatGPT visits to the sites it tracks relied on the tag alone, so phone app traffic was the most exposed.
Do UTM-tagged ChatGPT Ads clicks have the same problem?
Probably not. Search Engine Journal reported that Graphite's Ethan Smith said clicks from ChatGPT Ads carrying your own UTM parameters should be unaffected, because those tags are added by you rather than by ChatGPT. Relevant Audience still suggests advertisers check that their paid ChatGPT line held steady across 8 to 15 October. If you run ChatGPT Ads, compare paid sessions and conversions for that week with the previous fortnight, and confirm your campaign URLs still carry your own tags.
