Most of the traffic your site receives from AI platforms doesn't appear as AI traffic in your analytics. It appears as Direct. The platforms don't send referrer headers reliably, so GA4 classifies the sessions as unknown origin and groups them into the Direct channel — which already contains bookmarks, mobile apps, and typed URLs.
According to 1ClickReport's analysis, approximately 70% of AI referral traffic shows up as Direct in GA4 — adding to the broader picture of how AI affects organic traffic. This makes AI traffic genuinely difficult to measure — and means most sites are underestimating how much traffic AI platforms are sending.
The AI Traffic Market in 2025
For context on the scale: SE Ranking's 2025 AI traffic study found that AI-sourced referral traffic surged 527% year-over-year between January and May 2025 — from 17,076 to 107,100 monthly sessions across their tracked dataset.
ChatGPT sent 243.8 million visits to news and media sites in April 2025 alone — nearly double January's volume. But this AI traffic still represents only about 0.15% of global internet traffic versus 48.5% for organic search. It's growing fast but from a small base.
One genuinely useful data point from the SE Ranking study: AI visitors spend 68% more time on site than organic visitors — an average of 9 minutes 19 seconds versus 5 minutes 33 seconds. AI referral traffic is not just visitors; it's engaged visitors.
Why It Shows Up as Direct
When a user clicks a link in ChatGPT, the platform doesn't pass a referrer header in the standard way that websites do. This is a technical choice — some platforms do it deliberately for user privacy, others just don't implement referrer passing in their mobile apps or embedded browsers.
When GA4 receives a session with no referrer, it classifies it as Direct. This inflates your Direct channel with sessions that are actually AI-referred, and since Direct traffic already contains inherently ambiguous sessions, the problem is hard to separate out without additional instrumentation.
How to Track It More Accurately
Method 1: GA4 Channel Grouping
In July 2025, Google added an "AI assistant" channel group to GA4's documentation, recognizing AI platforms as a distinct traffic source for the first time. To use this: create a custom channel group in GA4 → Admin → Data Settings → Channel Groups → add a new channel with rules matching the known referrer strings from AI platforms.
Method 2: Referrer String Matching
For sessions that do pass referrer headers, GA4 will show the source domain. The known referrer strings for major platforms:
| Platform | Referrer String in GA4 | Notes |
|---|---|---|
| ChatGPT | chatgpt.com, chat.openai.com | Inconsistent across mobile/desktop |
| Perplexity | perplexity.ai | More consistent referrer passing |
| Google Gemini | gemini.google.com | Variable by interface |
| Claude (Anthropic) | claude.ai | Low volume currently |
Create a GA4 segment filtering for sessions where source contains any of these referrer strings. This captures the portion of AI traffic that does pass referrer headers — typically 30% of total AI referrals based on the 70% dark traffic figure.
Method 3: UTM Parameters on Cited Content
If you know specific pieces of content are being cited in AI responses (you can check this manually by querying the relevant topics in ChatGPT or Perplexity), you can add UTM parameters to those URLs if they're linked from external sources. This is a manual, targeted approach rather than a systematic one.
Method 4: Direct Channel Behavioral Analysis
AI referral sessions have distinct behavioral characteristics that can help identify them within your Direct channel: longer session duration (9+ minutes average), lower bounce rates, and typically landing on deep content pages rather than homepages. Segmenting your Direct channel by landing page depth and session duration can give a rough estimate of AI-referred volume within it.
What GSC Shows (and Doesn't)
Google Search Console shows only Google search traffic. It has no visibility into AI chat platform referrals at all. If you're trying to measure the full picture of how AI platforms affect your traffic — both the clicks they send directly and the clicks they reduce from Google Search via AI Overviews — you need GA4 alongside GSC.
This is an important distinction: AI Overviews reduce your GSC click data (because they absorb clicks that would have gone to organic results). AI chat platforms send referral traffic that shows up in GA4 (as Direct or with partial referrer attribution). These are two separate effects going in opposite directions.
Whether to Optimize for AI Visibility
The question following from all of this: should you actively try to get cited in AI responses? The honest answer is that the levers for AI citation are not fully understood yet, and what works in one AI system doesn't necessarily work in another.
What seems to correlate with AI citation across platforms: clear factual claims with specific numbers, authoritative domain signals, content structured around specific questions with direct answers, and external links from other authoritative sources pointing to the specific content. These overlap substantially with traditional E-E-A-T signals for Google Search.
The starting point is simply knowing whether AI platforms are sending you traffic at all, and what those visitors are doing when they arrive. Most sites don't yet have this measurement set up. Setting it up takes about 30 minutes in GA4 and immediately reveals data that's currently sitting invisibly in the Direct channel.