A drop in reported ChatGPT referrals does not necessarily mean fewer people are finding your brand. Some links in ChatGPT answers have recently appeared without the utm_source=chatgpt.com tag that helps analytics tools identify the source of a visit. If a click also arrives without a referrer, a tool such as Google Analytics 4 (GA4) may classify it as Direct.
This is a measurement problem, not proof that ChatGPT has stopped recommending your pages. It also does not establish that every ChatGPT link has lost its tag. Here is what has been observed, what remains uncertain, and how to audit your own data.
What changed with ChatGPT links?
UTM parameters are labels added to a URL. The utm_source parameter names the source that sent a visitor. A tagged link to your site might end with ?utm_source=chatgpt.com, making the visit easier to identify in analytics.
In research published on 14 October 2026, Profound found that the share of collected ChatGPT answers with at least one tagged cited link fell from about 98% on 8 October to 75% on 9 October and 66% on 14 October. Over the same period, answers using ChatGPT's new Intelligent UI became more common. OpenAI announced that interface on 7 October.
The timing suggests a connection, but it does not establish an announced policy to remove tracking. Profound also found tagged links in some additional panels even when a link shown directly in an answer lacked the tag. Which link a visitor clicks matters.
When both the UTM tag and browser referrer are absent, the destination site cannot identify ChatGPT as the source from that visit alone. Google describes (direct) / (none) as traffic without a clear referral source. Direct therefore includes more than people who typed your URL.
Why this matters for AI traffic attribution
Your site can still receive a visit, signup, or purchase even when the source label is missing. What changes is your ability to credit that action to ChatGPT. A fall in identifiable ChatGPT traffic could reflect fewer clicks, missing tracking information, or both.
Treat these as separate questions:
| Question | Useful signal | Limit |
|---|---|---|
| Is our brand appearing in AI answers? | Mentions and citations across a consistent set of prompts | An appearance does not prove a visit |
| Are people reaching our site from ChatGPT? | Tagged URLs and available referrers | Untagged, no-referrer clicks cannot be counted exactly |
| Are visitors taking valuable actions? | Landing-page visits, signups, leads, and purchases | A conversion without a source signal cannot be credited to ChatGPT with certainty |
AI visibility and AI traffic attribution complement each other. Visibility describes whether your brand appears in relevant answers. Attribution connects identifiable visits to what people do on your site. Neither metric can substitute for the other.
How to audit a suspected drop in GA4
Start with the date your ChatGPT referrals changed, then compare the same period across several views:
- Check identifiable ChatGPT sessions. In GA4's Traffic acquisition report, review session source and medium for
chatgpt.comand any ChatGPT tags you have historically received. - Check Direct sessions on the same landing pages. Look for a change around the same date, especially on pages that previously received ChatGPT visits. A matching movement is a clue, not proof that every Direct visit came from ChatGPT.
- Check total visits and outcomes. Compare page views, engagement, signups, and purchases. This helps distinguish a reporting change from a broader decline, although other campaigns or seasonality can also affect totals.
- Check the links themselves. Where you can reproduce the answer, inspect the destination URL and the referrer that reaches your site. Test different answer formats and devices; a tagged source link in one view does not establish that every link behaves the same way.
- Annotate the reporting break. Tell anyone comparing periods that identifiable ChatGPT referrals may be undercounted. Avoid treating the post-change number as directly comparable with earlier weeks.
Server logs can confirm visits and show any UTM parameters or referrer values that arrived. They cannot recover a source that the browser never sent. Similarly, an increase in Direct traffic is useful context, not a reliable count of hidden ChatGPT clicks.
Measure visibility alongside visits
People can encounter your brand in an AI answer without clicking through. Track mentions, cited pages, and share of voice for a stable set of relevant prompts alongside your traffic and conversion reports. Keep the prompt set and measurement method consistent so changes are interpretable.
PingAura's AI visibility monitoring helps teams follow mentions and citations across AI answers. Its AI Traffic view can report visits with identifiable AI source signals. When a click arrives with neither a UTM tag nor a referrer, PingAura cannot definitively attribute that visit to ChatGPT either. Use the visibility and traffic views together to understand what each can actually measure.
FAQs
Did ChatGPT remove UTM parameters from every link?
No such blanket change has been established. October 2026 research found missing tags on many links in a newer answer format, while some links in other panels still carried tags. Check the specific link and surface before drawing a conclusion.
Why might a ChatGPT visit appear as Direct in GA4?
If the landing URL has no source tag and the browser sends no referrer, GA4 has no clear source for that visit. Google groups visits without clear referral information under (direct) / (none). Other causes of Direct traffic still exist.
Can a rise in Direct traffic reveal the missing ChatGPT clicks?
It can suggest a measurement shift when it coincides with a fall in identifiable ChatGPT referrals on the same pages. It cannot tell you exactly how many Direct visits came from ChatGPT.
Can server logs restore attribution?
Logs show the URL, query parameters, and referrer received by your server. If both source signals are absent, a log cannot identify the originating ChatGPT click on its own.
Does this change how to optimise for AI answers?
The missing tags affect measurement of some clicks. They do not demonstrate a change in which content ChatGPT selects or cites. Continue measuring answer visibility and improving useful, accurate content while you investigate the traffic report.
What to do now
Audit your own GA4 data and a sample of live links before changing your AEO strategy. Mark the point where tracking may have changed, compare identifiable referrals with Direct and total visits, and keep citations and brand mentions in a separate visibility report. That gives your team an honest view of what is measured and what remains unknown.
