ChatGPT Traffic Glitch: What Happened and How to Track It

For about a week in October 2026, your analytics showed ChatGPT traffic to your site dropping, but that isn’t what happened.

After OpenAI rolled out GPT-6 in early October 2026, links in ChatGPT answers stopped carrying the ChatGPT UTM tag (utm_source=chatgpt.com) and stopped passing referrer data with the click. Those visits were filed under “Direct” instead. OpenAI has since started fixing the problem, but for brands that had finally started to see ChatGPT as a line item in analytics, the week-long blind spot is a warning.

We’ve been saying for a while that clicks were never the best way to measure how AI search treats your brand, and this glitch proved it. The real question is whether you’re mentioned, recommended, and cited in the answer itself. That’s what our AI Visibility Scan was built to measure.

What Happened to ChatGPT Traffic in October

On October 14th, Search Engine Roundtable reported that ChatGPT links had stopped carrying UTM parameters, starting with the GPT-6 release about a week earlier.

The diagnosis came from Cedric Huesler of Adobe: GPT-6 was opening outbound links with noopener,noreferrer and dropping the utm_source=chatgpt.com tag. With no tag and no referrer, analytics platforms had nothing to attribute the visit to, so it landed in the Direct traffic channel. Across roughly 500 pages, he found ChatGPT referrals down about 31,000 visits a day while Direct rose about 26,000 a day on the same pages.

On October 15, Huesler updated his post: OpenAI had confirmed it was a bug, and ChatGPT was tagging its links with utm_source=chatgpt.com again. OpenAI hasn’t published an official notice or release note about it. Huesler also noted that the referrer is still missing on most ChatGPT clicks, so the UTM tag is now the signal to count.

Why the Glitch Still Matters for ChatGPT Traffic Reporting

If your team uses ChatGPT traffic to justify AI search investment, your October reports will show a dip for roughly October 8–15, even though actual visibility didn’t change. That creates three problems:

• False negatives, because a brand could be cited more often than ever and still watch “ChatGPT traffic” fall off a cliff.
• Inflated Direct traffic for that window, which carried a hidden AI component and muddies every channel comparison that relies on it.
• Harder budget conversations with stakeholders who only look at sessions and will see AI search contributing less, right when it matters more.

Analytics can be corrected after the fact. But the episode showed how much AI search reporting hangs on a single tag, and that tag can’t tell you whether AI engines are recommending you in the first place.

Clicks Were Never the Whole Story

Even when the UTM tag worked, referral traffic captured only a sliver of AI search’s influence. Most AI answers never produce a click. A buyer asks ChatGPT for the best option in your category, reads a synthesized shortlist with a one-line verdict on each brand, and either remembers you or doesn’t.

That’s the idea behind AI visibility – how often, how prominently, and how favorably AI assistants like ChatGPT, Claude, Gemini, and Perplexity mention your brand when people ask questions in your category. In AI answers there is no page two, you’re either in the consideration set with a position and sentiment attached, or you’re invisible for that question.

Clicks vs. Visibility: Why measuring citations and direct AI answers matters more than relying on traditional referral clicks.

How Gen3’s AI Visibility Scan Fills the Gap

The AI Visibility Scan measures AI search at the source: the answer, not the click. Gen3 launched its AI Visibility tool in August 2026 because every tool we tried for clients answered “are we mentioned?” and none answered “what should we change?” We built one that does both, and it doesn’t depend on a referral tag surviving the next LLM update.

The scan runs hundreds of realistic buyer questions across ChatGPT, Gemini, Perplexity, and Claude, repeats each prompt across multiple runs to account for answer variance, and analyzes every response. Here’s how that maps to what you just lost in analytics:

Every scan rolls up into a 0–100 composite score covering mentions, sentiment, citations, and accuracy: one trend line for leadership that doesn’t swing every time a platform changes how it opens links.

The next time your ChatGPT traffic line drops, you’ll be able to tell stakeholders whether your AI visibility actually fell, or whether the measurement just broke.

What to Do Now

• Annotate your analytics. Mark October 8–15, 2026 in GA4, Adobe Analytics, or your platform of choice so nobody mistakes a tracking glitch for a performance drop in MoM or YoY reporting.
• Confirm the rebound. To check traffic from ChatGPT, filter on the utm_source=chatgpt.com tag rather than the referrer, which is still missing on most ChatGPT clicks. Tagged sessions should recover from October 15 onward. Compare them with Direct traffic on your top ChatGPT landing pages for the same dates.
• Explain the dip before anyone asks. Tell leadership, and clients if you’re an agency, that October’s ChatGPT numbers include a week of misattributed traffic. Get there before the October report does.
• Make sure ChatGPT can see your site. OpenAI’s crawlers don’t execute JavaScript, so client-rendered pages can look empty to them. Allow OAI-SearchBot in robots.txt, server-render key pages, and confirm Bing indexation. Our guide on how to check if your brand shows up in ChatGPT walks through the full audit.
• Measure visibility at the source. Set up recurring AI visibility scans for a stable cross-platform baseline of mentions, sentiment, citations, and share of voice that doesn’t depend on how any one platform handles links.

A Note for SEO and GEO Professionals

Generative engine optimization (GEO) is the practice of improving how often and how favorably AI assistants mention and cite a brand in their answers. For practitioners, this change is less about a lost metric and more about how AI search work gets reported and defended:

• Demote ChatGPT referrals as a primary GEO KPI. One glitch knocked the metric out for a week. Keep it as a supporting metric with a footnote, and lead client scorecards with mention rate, share of voice, answer position, and citation rate.
• Shift attribution toward citations. Owned citations, meaning AI answers that cite pages on your client’s domain, are now the closest proxy for “AI sent us this visitor.” Track which URLs get cited and tie content work to those pages.
• Watch server logs, not just analytics. AI crawler activity (OAI-SearchBot, ChatGPT-User, GPTBot) still shows up in logs and reveals which pages AI engines fetch, even when the resulting clicks go unattributed.

Measure What AI Says, Not What It Leaks

The ChatGPT glitch was a measurement problem, not a visibility problem. Your brand was still being discussed in AI answers the entire week, analytics just couldn’t see the clicks.

This one got fixed in about a week, but the next one might not. The brands that come out ahead will measure what AI engines actually say, not the traffic those engines leak. As we put it when we first introduced AI brand visibility monitoring: AI is already talking about your brand. The question is whether you know what it’s saying.

Questions Marketers Are Asking About ChatGPT Traffic

Has ChatGPT usage gone down, or was it a tracking issue?

The October 2026 drop in reported ChatGPT traffic was a tracking glitch, not a sign of lower usage. For about a week after the GPT-6 release, ChatGPT clicks arrived without the utm_source=chatgpt.com tag or a referrer, and Huesler’s data showed the lost ChatGPT visits reappearing as Direct traffic on the same pages. OpenAI began fixing it on October 15.

How do I track traffic from ChatGPT now?

The UTM source ChatGPT tag (utm_source=chatgpt.com) is back on ChatGPT links, so filter for it in GA4 or your analytics platform rather than relying on the referrer, which is still missing on most ChatGPT clicks. For the October 8–15 gap, compare Direct traffic on pages that historically drew ChatGPT visits to estimate what was misattributed.

Does a drop in ChatGPT referral traffic mean my AI visibility dropped?

Not necessarily. Referral data measures clicks, not visibility. Your brand can be mentioned and cited as often as before while reported ChatGPT traffic falls, as it did in October. Measuring the answers directly is the only way to know.

What tools track ChatGPT and Perplexity traffic?

Traffic analytics tools such as GA4 and Adobe Analytics can only attribute AI visits when a platform passes a referrer or UTM tag, and as October showed, that tagging can break without warning. To measure ChatGPT and Perplexity side by side, pair analytics with an AI visibility tool such as Gen3’s AI Visibility Scan, which tracks mentions, citations, and sentiment across ChatGPT, Perplexity, Gemini, and Claude.

See where you stand. Run a free Pulse visibility check to see how major AI assistants describe your brand when buyers ask real questions in your category. No account required. Or talk to Gen3’s SEO & AI Search (GEO) team about building AI visibility into your reporting.