ChatGPT stopped tagging outbound links on Oct 9, inflating GA4 Direct

Ethan Smith, writing on LinkedIn on 14 October, reports that ChatGPT-attributed organic sessions in GA4 fell by roughly 45% across several sites he tracks. The drops on individual sites ranged from 22% to 57%, per Smith. Google and Direct stayed flat or grew over the same window.

The first dip shows on 8 October, with a larger step on 9 October. Smith attributes this to ChatGPT opening outbound links with noreferrer and without utm_source=chatgpt.com. GA4 has no source to read, so the visit lands in Direct.

Smith notes the Sources panel and shopping links still carry the UTM. Inline links do not. He says 95% of mobile ChatGPT visits relied on the UTM alone, which makes app traffic the most exposed surface.

Tomek Rudzki of Peec AI, cited in the same post, reports that the new ChatGPT model links straight to sites four to five times more often, with a bias toward homepages. Those homepage links are untagged.

Reading the gap in your own GA4

Annotate 8 October in your reporting. Smith's view is that ChatGPT numbers from that date forward are unreliable until OpenAI ships a fix, which he expects soon.

Smith suggests looking at ChatGPT sessions and homepage Direct as one picture. On one site he reviewed, homepage Direct rose 17% while Google was flat, and the visitors were all new. New-user Direct on pages that previously drew ChatGPT traffic is the pattern to watch.

ChatGPT Ads clicks that carry your own UTMs are unaffected, per Smith. Verify the pattern on your own property before you change a dashboard or brief a client.

A measurement path that does not depend on OpenAI

A post-signup "how did you hear about us" question gives you an attribution source that OpenAI cannot silently break. We think the current gap in GA4 is a reasonable moment to stand one up, alongside the server-side signals you already trust. Treat the GA4 Direct inflation as one input while you rebuild the picture from sources you control.