Why GA4 shows so much direct / none traffic, including ChatGPT and Perplexity visits

GA4 puts a visit in (direct) / (none) when it finds no clear referral source, such as a link without UTM parameters. AI assistants are different. In 9 of the 29 GA4 properties we manage (July 30 to October 28, 2026), ChatGPT, Gemini, Claude and Perplexity visits landed in GA4’s AI Assistant channel, and also under other mediums.

What did we find in the GA4 properties we manage?

Our own data, not a vendor’s.

We queried the GA4 properties we manage for sessions whose source names an AI assistant, over the 90 days from July 30 to October 28, 2026. AI-assistant visits showed up in 9 of the 29.

In those properties, ChatGPT arrives as chatgpt.com / ai-assistant and Perplexity as perplexity.ai / ai-assistant, alongside claude.ai and gemini.google.com with the same medium.

The same assistants also arrive other ways. We see chatgpt.com / referral, organic, feed and (none), and bare source names such as perplexity / (not set) and gemini / (not set). We haven’t established why those visits differ, so we won’t guess. The practical point is that GA4’s AI Assistant channel is where to start, not the whole answer.

What does (direct) / (none) mean in GA4?

Google defines it as “website traffic that doesn’t have a clear referral source.” In GA4’s default channel group, a session lands in Direct when its source is exactly (direct) and its medium is (not set) or (none). That channel is meant for visitors who arrive “via a saved link or by entering your URL.”

So Direct is a real channel. It also collects visits whose source information was lost on the way in.

Why is GA4 showing so much direct / none traffic?

Because source information got lost before GA4 could record it. Google names these causes on its own help page.

• Missing UTM parameters. When links lack UTM parameters, or your site isn’t integrated with your marketing and advertising platforms, “traffic source information is lost.” Missing UTMs also produce (not set) values in reports.
• Redirects. Redirects can strip UTM parameters from the URL before GA4 sees them.
• Ad blockers. They can interfere with the tracking cookies GA4 uses to identify where traffic came from.

You can check the first two yourself. Click your own campaign links (the ones in your emails and ads) and watch what reaches the landing page.

What’s the difference between Direct and Unassigned?

Direct is a channel; Unassigned is GA4 saying no channel fit. Unassigned is the value GA4 uses “when there are no other channel rules that match the event data.” Direct, by contrast, has its own rule: source (direct), medium (not set) or (none).

How do I see ChatGPT and Perplexity traffic in GA4?

Open the AI Assistant channel in your Default Channel Group reports. Google added it on May 13, 2026 to show how users discover your site “through chatbots like ChatGPT, Gemini, and Claude.”

The rule is simple. When the referrer matches GA4’s list of recognized AI assistants, GA4 sets the medium to “ai-assistant” and the campaign to “(ai-assistant),” and the AI Assistant channel is every session whose medium exactly matches “ai-assistant.” To see which assistant sent a visit, look at session source / medium. In GA4’s Data API, sessionSource is “the source that initiated a session.”

Does the AI Assistant channel catch every AI visit?

Not in the properties we manage, as the finding above shows. If you want visits from these assistants in one place, build a custom channel group:

1. In Admin, go to Data display > Channel groups and click “Create new channel group.” That starts you with a copy of the default group.
2. Add a channel with one condition: Source, matches regex, .*(chatgpt|openai|perplexity|gemini|claude|copilot).*
3. Move it above Organic Search and Referral. Traffic “is included in the first channel whose definition it matches,” and Google’s own AI-assistant example puts the channel above Referrals and says it can also go above Organic Search. Then save.

Channel definitions aren’t case sensitive. A standard property gets 2 custom channel groups (a 360 property gets 5), with up to 50 channels in each.

One caveat before you trust the number. A pattern this broad also catches unrelated sites whose names contain those words, so review the list of sources it matches.

Will server-side tracking fix direct / none traffic?

Not on its own. Server-side tracking changes where data is processed, not where it originates. A click that arrives with no referrer and no UTM parameters is unattributed before any server touches it.

The fix for missing or stripped UTM parameters is in your links and redirects, not in your tracking infrastructure.

When don’t you need to worry about this?

If every campaign link you control carries UTM parameters, survives its redirects, and your Direct traffic looks like people typing your name, your Direct number may simply be accurate. The ad-blocker cause is the one that doesn’t go away with better links. It’s also the one where the Meta pixel loses data too.

Related

• → Why Meta ads under-report conversions on Shopify
• → Who can set up server-side tracking: DIY, app or managed agency
• → Stape alternatives, compared
• → Waterloo Track, our managed server-side tracking service

If your Direct number is too high to explain and you’d like the tracking pipeline handled for you, see what Waterloo Track covers.