ChatGPT Traffic in Google Analytics: What the Numbers Miss

If you check ChatGPT traffic in Google Analytics, you probably assume the figure is roughly right. In mid-October, for some sites it suddenly wasn’t. ChatGPT referrals fell sharply, “Direct” traffic rose by almost the same amount, and plenty of business owners will have read that as AI search drying up.

It wasn’t. It was a tracking change, and by most accounts a bug that was quickly fixed. But it showed how fragile AI referral data is, and how easily a wrong number turns into a wrong decision about where to spend your marketing budget.

This article explains what happened, where AI visits show up in GA4 today, and the checks we’d run before trusting any figure that claims to show how many enquiries ChatGPT sends you.

What happened to ChatGPT traffic in Google Analytics in October

Since mid-2025, ChatGPT has added a utm_source tag to many of the links it shows, which lets analytics tools credit the visit to ChatGPT (Search Engine Roundtable, June 2025). That tag, together with the referrer your browser normally passes on, is how GA4 knows a visitor came from an AI assistant.

Around 9 October 2026, shortly after OpenAI released GPT-6, Adobe’s Cedric Huesler reported that ChatGPT had started opening links with the tag stripped and no referrer passed. In his words, reported by Search Engine Roundtable: “There’s no tag and no referrer, so your analytics files the visit under direct.”

Across roughly 500 pages where ChatGPT normally drove at least 5% of visits, he found ChatGPT visits down by about 31,000 a day and Direct visits up by about 26,000 a day on the same pages, comparing 9–13 October with the week before (Search Engine Land). On 15 October he updated his post to say it was a bug and that ChatGPT was tagging links again. OpenAI hasn’t commented publicly.

The practical point: if you looked at your AI traffic for that week, the drop probably wasn’t real. If you compare October with September later this year, expect a dip that may be an accounting error, not a change in demand.

Where AI referral traffic shows up in GA4 now

The AI Assistant channel

In May 2026 Google added an AI Assistant channel to GA4’s default channel group. When a visit’s referrer matches a recognised AI assistant, such as ChatGPT, Gemini or Claude, GA4 sets the medium to “ai-assistant” and files the session under that channel (Google Analytics Help). You don’t have to set anything up.

That’s useful, but Google hasn’t published the full list of recognised referrers, so newer or smaller AI tools may still appear under Referral (Search Engine Journal).

Why some AI visits still land in Direct

The channel depends on a referrer or a tag. Remove both, as happened in October, and GA4 has nothing to go on. The same applies when someone copies a link out of a chat into their browser, or taps through from an app that doesn’t pass referrer data. Those visits are counted as Direct.

So treat your AI Assistant figure as a floor, not a total. Some AI-driven visits are always hiding elsewhere.

If you’re not sure which of your channels actually produce enquiries, our free health check includes a manual review of your SEO, PPC, website and conversion set-up.

How to check your own ChatGPT traffic in Google Analytics

This takes about ten minutes and needs no technical skill.

1. Find the channel. In GA4 go to Reports, then Acquisition, then Traffic acquisition. With the primary dimension set to the session default channel group, look for “AI Assistant”.
2. See which assistants. Switch the dimension to session source / medium and look for entries such as chatgpt.com, gemini.google.com or claude.ai.
3. Look at the pages they land on. Add landing page as a secondary dimension. AI visitors often arrive on a specific service page or guide rather than your homepage. That tells you which content AI tools are citing.
4. Check what they do next. Compare engagement rate and key events (form submissions, calls, purchases) for AI Assistant against Organic Search. A small channel that converts well deserves more attention than a big one that doesn’t.
5. Mark the anomaly. Add a note in your own reporting for 9–15 October 2026 so nobody misreads that week in six months’ time.

If step 4 shows nothing at all, the problem may not be AI. It may be that your key events aren’t set up properly, which is far more common than most business owners realise.

Why is direct traffic increasing, and when should you worry?

Direct is GA4’s “we don’t know” bucket. A rise can mean more people typing your address, but it can just as easily mean broken tagging, a consent banner change or an AI tool dropping its referrer. Some simple decision rules:

• Direct up, AI Assistant or Referral down by a similar amount, same dates: almost certainly an attribution change. Don’t change your strategy over it.
• Direct up on deep pages (a specific blog post or service page) rather than the homepage: suspicious. Few people type a long URL from memory. Those visits are likely coming from links that lost their source.
• Direct and Organic Search both down: look wider. Google ran a spam update from 24 September to 8 October, and site owners reported further ranking volatility from 16 October. Check Search Console before blaming your analytics.
• Traffic stable but enquiries down: the issue is on your website, not in your reporting. That’s a conversion problem.

Measure enquiries, not just visits

Here’s our opinion: for most SMEs, the exact number of ChatGPT visits matters far less than whether AI-referred people become customers. Visits are easy to lose in the plumbing. Enquiries are harder to lose if you track them properly.

Take an illustrative example. A Burnley kitchen fitter gets most of their leads by phone. GA4 shows a few dozen AI Assistant sessions a month and concludes it’s a rounding error. But if those visitors read the pricing page and then ring the number on screen, GA4 records a visit and no conversion. The channel looks worthless when it may be one of the best.

The fixes are unglamorous and they work:

• Track phone clicks and form submissions as key events, so every channel, AI included, is judged on leads.
• Ask “How did you hear about us?” on your enquiry form and on the phone. Add “ChatGPT or another AI tool” as an option. Self-reported data catches what analytics misses.
• Pass the source into your CRM with a hidden form field, so you can see which leads turned into paying work.
• Make the pages AI tools cite convert. If ChatGPT sends people to one guide, make sure that page has a clear next step, not just a link to your homepage. This is where conversion rate optimisation earns its keep.

For a Manchester e-commerce brand (again, illustrative), the equivalent check is revenue by channel. If AI Assistant sessions show a healthy average order value, that’s a reason to keep product information clear and accurate wherever AI tools pick it up.

Getting this plumbing right is a large part of what our digital consultancy work covers: tracking, CRM links and reporting that a business owner can trust.

Should you chase AI referral traffic at all?

Yes, but in proportion. Check your own figures first: for many local service businesses, Google search will still be the far bigger source of visits. The argument for paying attention is that someone arriving from an AI answer has often already done their research. Test whether that holds in your conversion data before you invest heavily.

The way to win more of those visits is mostly good SEO done properly: clear service pages, answers to real customer questions, accurate business information and a technically sound site. We covered this in more depth in our guide to AI search optimisation, and it’s built into our SEO services.

What you shouldn’t do is rebuild your strategy around one week’s data from a channel whose tracking changes without notice. October proved that.

How MonkeyFish Group approaches AI traffic measurement

When we review a client’s analytics, we start with the leads, not the traffic. We check that calls, forms and sales are recorded as key events, then work backwards to see which channels produce them.

We then look for the leaks: Direct traffic landing on deep pages, referrers that should be grouped as AI, consent settings that hide visits, and enquiry forms that don’t record a source. Finally we set up reporting that flags anomalies like October’s, so a tracking bug is spotted as a bug and not mistaken for a trend.

It’s practical work, done by people with 20+ years’ experience of watching measurement tools change under their feet, for businesses across Burnley, Lancashire and the North West.