Seeing further, standing on giants

Start a Conversation
By: /

Did ChatGPT Traffic Drop? Why AI Referral Traffic Hid in Direct, and How to Track It

In mid-October, plenty of marketers opened Google Analytics and did a double take: traffic from ChatGPT had fallen by a third to a half in a single week. Had AI stopped recommending them? It turned out the answer was no. Those visitors were there all along; they had simply been recorded as "direct" traffic. The episode is a useful reminder that AI referral traffic tracking is more fragile than it looks, and that brands should strengthen how they measure it now rather than react to a single number.

What Happened: ChatGPT's Tracking Tag Disappeared After GPT-6

A digital marketing manager frowning at a laptop where a website traffic chart shows a sudden steep drop

Pieced together from several analysts, the sequence looks like this:

  • Around October 8: After GPT-6 rolled into ChatGPT, links in ChatGPT answers stopped carrying the utm_source=chatgpt.com tag. Since most ChatGPT clicks already arrive without a referrer, analytics tools could no longer tell where the visitors came from and filed them under direct traffic.
  • How big: Adobe's Cedric Huesler analyzed about 500 pages where ChatGPT supplied at least 5% of visits. From October 9 to 13, ChatGPT visits fell by about 31,000 a day, while direct visits to the same pages rose by about 26,000 a day. AI search analytics firm Profound saw the share of ChatGPT answers with a tagged link fall from about 98% on October 8 to 66% on October 14.
  • Fixed on October 15: Huesler said OpenAI confirmed to him that this was a known bug. A fix began rolling out early on the 15th, and by that morning tagged clicks were back to 95–100% of the same hour a week earlier. OpenAI has not commented publicly.

In other words, the traffic didn't vanish. It just changed its name in your reports.

The Weak Spot in AI Referral Traffic: A Tag That Can Be Taken Away

A hand tapping a link card in an AI chat on a smartphone, with a web analytics dashboard on a laptop behind it

What matters here is not one week of missing data but the structural issue underneath:

  • A tag is a courtesy, not an obligation. Whether an AI assistant adds a source tag or passes a referrer is entirely the platform's call. It can change at any time, without notice.
  • The referrer is still mostly missing. The tag is back, but most ChatGPT clicks still carry no referrer, so the tag is almost the only clue to the source.
  • Misrecorded data stays misrecorded. Visits logged as direct between October 8 and 15 will not be reassigned later.
  • Not all links are treated the same. Analysts observed that inline answer links lost the tag, while links in the sources panel and shopping results largely kept it.

If your company decides budgets or content direction based on "how much traffic ChatGPT sends," an invisible change like this can easily lead to the wrong conclusion.

Four Things You Can Do Now

1. Annotate October 8–15 in Your Reports

Add a note in Google Analytics or your analytics tool flagging that AI traffic was undercounted during this window. That way monthly, quarterly and year-over-year comparisons won't suggest your AI traffic suddenly halved. And don't rebuild attribution models around the dip.

2. Look at AI Sources and Direct Traffic Together

Create a dedicated channel group for AI sources and watch direct traffic to the same landing pages alongside it. If AI traffic drops while direct traffic rises at the same time, suspect tracking before you suspect your marketing.

3. Use Your Own Tags Wherever You Control the Link

Wherever you place the link yourself, such as ChatGPT ads, partner articles and social posts, add your own UTM parameters. Analysts noted that ChatGPT ad clicks carrying advertisers' own UTMs may have been unaffected this time.

4. Look Beyond Clicks to Brand Search and Conversions

Many people see a brand in an AI answer and don't click; they search for it or type the URL later. Track branded search volume, conversions from direct traffic and a simple "How did you hear about us?" question to see AI's full impact. For a deeper look at reading AI visibility tools, see Same Brand, Different Scores: How to Read AI Visibility Measurement With the IAB's Four Ps.

Beyond the Numbers: Being Worth Recommending Is What Counts

A brand manager interviewing a product expert at a studio table with product samples while a colleague films on a tripod

Tracking tools break and change, but one thing doesn't: AI only recommends content it can find and trust. Rather than being pulled around by every swing in the numbers, brands should keep investing in:

  • Clear, quotable content: State product specs, prices and service scope plainly so AI can relay them accurately.
  • Named expert voices: Explanations signed by real experts earn more trust than anonymous marketing copy. Read more in Who Said It Matters More: Named Experts Nearly Double Trust in AI Answers.
  • Landing pages you control: Make sure visitors arriving from AI see the answer they came for the moment they land.

Data is there to inform decisions, not to frighten you. AI keeps the numbers straight; people decide what they mean.

TitApeX helps brands build websites, tracking setups and content strategies that are visible in the AI era and measurable too. If you'd like a check-up of your site's AI referral traffic tracking, explore our services or get in touch.

References

Leave a Reply

Your email address will not be published. Required fields are marked *