Why does traffic from ChatGPT show up as direct in analytics?

Analytics tools sort a visit by two clues: a campaign tag in the URL and the referrer the browser sends. If both are missing, the visit has no recorded origin, and the tool calls it direct.

ChatGPT supplies the first clue on purpose. OpenAI's publisher FAQ says ChatGPT automatically adds the UTM parameter utm_source=chatgpt.com to referral URLs. A visit carrying that tag shows up under its source, whatever the browser does with the referrer.

The tag doesn't travel on every click. As of October 2026, clicks from the free tier and the mobile app frequently arrive untagged. If the referrer is also dropped on the way, the visit lands in direct with nothing to show it came from ChatGPT.

So your direct bucket holds a slice of ChatGPT traffic you can't see, and nobody has published how big that slice is. The next move is to stop reading the ChatGPT line in your reports as a total. It is the tagged portion only. Our guide to measuring AI referral traffic covers the full setup.

Not all AI traffic fails the same way. This table sorts the common cases.

The last row is a different problem. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they never fire a tag-based analytics script. That's covered in why AI crawler visits don't show up in Google Analytics.

The fourth row matters more than it looks. ChatGPT doesn't search for every question. When it answers from what the model already knows, it can name a brand with no link at all. A reader who acts on that answer arrives as direct or as a branded search, and no tag can catch it.

The firm facts are short. OpenAI documents the utm_source=chatgpt.com tag. The tag is known to be missing on some surfaces. Google AI Overviews emit no referral event.

That last fact skews any comparison between platforms. ChatGPT stamps its links, while most rival AI products don't. A report that shows ChatGPT sending nearly all of your AI referrals is partly measuring which platforms tag their links. That's why claims about ChatGPT's share of AI referral traffic are upper bounds for ChatGPT's tracked share, never estimates of real usage.

Google's own reporting doesn't close the gap. As of October 2026, the Search Console report for AI Overviews and AI Mode shows impressions only, with no clicks, no click-through rate and no position. Google also says the two surfaces may use different models and techniques and show different links. You can see that Google displayed your page in an AI feature. You can't see whether anyone clicked.

What nobody publishes: the share of ChatGPT clicks that arrive untagged, or the share of direct traffic that is really AI. Any tool quoting a precise number for either is estimating.

A few behaviours push AI visits into the wrong bucket even when the platform tags its links.

• Copy and paste. A visitor who copies a link from an answer into a new tab may strip the tag, or keep it, depending on what they copied.
• In-app and default browsers. Links opened from a phone app can hand off to another browser, which may not pass a referrer.
• Shared answers. A colleague forwards a ChatGPT answer in chat or email. The click carries the tag if the link kept it, but the visitor never used ChatGPT.
• Delayed visits. Someone reads an answer at lunch and visits from a laptop that evening. That's direct or branded search, with no link back to the answer.
• Changing analytics defaults. Google Analytics added a native AI assistant channel in May 2026. Your channel numbers can shift when the definition changes, not just when traffic does.

None of these are fixable from your side. They're reasons to treat the AI line as a floor and to look for the rest elsewhere.

You can't recover the missing tags. You can build a better picture around them.

1. Group every labelled AI source in one channel. Match utm_source=chatgpt.com and known AI referrer domains. The free AI referral tracking setup walks through this.
2. Watch direct traffic to deep pages. People rarely type a long product or article URL by hand. A rise in direct visits to deep pages with no campaign behind it is a signal worth checking, not a proof.
3. Ask people. Add a "how did you hear about us" field to signup and contact forms, with an AI assistant option. Self-reported answers catch the unlinked mentions no tag can.
4. Read the platform reports separately. Search Console shows AI Overviews and AI Mode impressions. Bing Webmaster Tools has an AI Performance report for Copilot citations. Neither talks to your analytics.
5. Measure the answers, not just the clicks. Referral data tells you who clicked. It doesn't tell you whether an AI names you at all. The free checker on this site shows whether AI engines name you for up to three buyer questions, as a one-off snapshot.

If you need to tie revenue to these visits, read whether you can attribute leads and sales to AI search before you promise anyone a number.