ChatGPT traffic may be showing up in your analytics under the wrong label. Search Engine Roundtable’s Barry Schwartz reported on October 14, 2026 that links in ChatGPT answers appear to have lost their UTM tagging after OpenAI shipped GPT-6, roughly seven days before his post. The practical result is that real AI referrals look like direct visits.
The mechanism comes from Cedric Huesler of Adobe, who laid it out on LinkedIn. He says GPT-6 opens outbound links in a new tab with the noopener and noreferrer options set, and no longer appends utm_source=chatgpt.com. With neither a tag nor a referrer header, an analytics platform has nothing to attribute the visit to, so it defaults to direct. Schwartz notes that OpenAI began tagging ChatGPT outbound links with UTM data in June, so October marks a rollback. Schwartz said he is unsure whether the change was deliberate, and OpenAI has not commented in the source.
Huesler backed the claim with a dataset. He examined roughly 500 pages where ChatGPT supplied at least 5 percent of visits, comparing October 9 to 13 against the preceding week. ChatGPT-attributed visits fell by 31,000 a day on those pages, while direct visits rose by 26,000 a day. Search dropped 7 percent, which he described as the same decline seen across the board, so it does not explain the shift.
The page-level pattern is the more persuasive part. Direct traffic rose on two of every three pages that lost ChatGPT visits. In the prior week the direction was effectively random. Search on those same pages grew on just 45 percent of them. A general traffic wobble would not be expected to line up so neatly with ChatGPT’s losses.
Huesler also checked who the new direct visitors are. On mobile, the share of direct visitors who saw the consent banner, a stand-in for being new to the site, climbed from 36 percent to 44 percent. Among ChatGPT visitors that share is 55 percent. A reasonable reading is that the direct channel now contains a blend of ordinary direct traffic and relabeled ChatGPT visitors, though Huesler’s figures do not prove the proportions.
Treat the evidence with care. It comes from one analyst’s sample, built from pages that already depended heavily on ChatGPT, so sites with little AI referral traffic may see nothing. The 31,000 daily drop also exceeds the 26,000 daily gain, and the source does not explain the remaining difference.
A second report points to a separate signal. Sébastien Billard wrote on X on October 14 that he has seen a sharp fall in human ChatGPT traffic since October 9, plus a drop in ChatGPT-User bot traffic. His logs show OAI Search bot activity rising since October 7. He said he would look into the UTM parameter. Bot-side movements are not obviously caused by a referrer change, so they should be analyzed on their own.
For analytics teams, five checks are worth running now:
- Chart ChatGPT-attributed sessions against direct sessions by landing page from October 9 onward, and look for mirrored moves.
- Segment direct traffic by landing page. Deep article URLs rarely attract genuine typed-in visits, so direct growth there is suspect.
- Compare new-visitor share on direct traffic before and after October 9, as Huesler did with the consent banner.
- Inspect server logs for missing referrer headers and stripped query strings.
- Annotate dashboards with the date so stakeholders do not read a ChatGPT decline as lost demand.
Until OpenAI says whether this is intentional, treat unexplained direct growth on pages that formerly earned ChatGPT visits as probable AI referral traffic, and report it as a separate estimate rather than announcing a collapse.
Search Engine Roundtable, Barry Schwartz, October 14, 2026, citing Cedric Huesler of Adobe on LinkedIn.
