Quick answer: As of May 13, 2026, GA4 added a default channel called "AI Assistant" that captures referrals from ChatGPT, Gemini, Claude, Perplexity, and a handful of other AI tools. It catches most paid-tier traffic but misses traffic from free-tier ChatGPT (which strips the referrer) and any AI tool not on Google's recognized list. To audit your true AI traffic, you need a hybrid approach: native channel + UTM tagging on AI-exposed content + landing page heuristics.
What changed on May 13, 2026
Google released an update to the default channel grouping that adds AI Assistant as a primary channel between Organic Search and Referral. It triggers when the session referrer matches a maintained list of AI assistant domains: chat.openai.com, chatgpt.com, gemini.google.com, claude.ai, perplexity.ai, copilot.microsoft.com, and others.
Before this change, AI-referred traffic landed in Referral (when the referrer was preserved) or Direct (when it wasn't). The new channel makes AI attribution visible in default reports without custom channel groups.
What the new channel captures
- Paid-tier ChatGPT clicks (referrer preserved)
- Gemini browse-the-web responses with citation links
- Claude responses with web citations (Claude 3.5+ models)
- Perplexity citation clicks (already had clear referrer)
- Bing Copilot citation clicks
What it still misses
Free-tier ChatGPT
OpenAI strips referrer information for free-tier users when they click a link from a chat response. These users land on your site as direct/(none) — the new AI Assistant channel can't catch them because there's no referrer to read.
Google AI Overviews (AIO)
The AI-generated overview that appears at the top of Google SERPs sends clicks as google/organic, not as AI Assistant. There's no way to distinguish AIO clicks from regular organic clicks without query-level data from Search Console.
In-app AI assistants
Notion AI, Slack AI, Microsoft 365 Copilot in desktop apps — all of these strip referrers because the user is in a native application, not a browser. Traffic from these lands in Direct.
Voice assistants
Alexa, Google Assistant, Siri — when these surface a web result and the user taps through, the referrer is usually stripped or set to a meaningless app identifier. Not captured.
The audit: how to measure your real AI traffic
Step 1: Compare AI Assistant vs prior Direct baseline
Pull the last 90 days of Direct traffic from before May 13, 2026. Compare to the same period after May 13. If Direct dropped by ~10–30% and AI Assistant absorbed the difference, the new channel is correctly capturing AI referrals that were previously misattributed.
Step 2: Audit your landing pages for AI exposure
Some pages are more likely to be cited by AI assistants than others. To estimate your true AI-driven traffic (including the unattributed portion):
- Identify pages that rank for question-style queries (use Search Console → Performance → Queries → filter "how", "what", "why")
- For each, check: AI Assistant + Direct sessions over the last 30 days
- Compare the Direct % on these pages to your site-wide Direct %. If significantly higher, the excess is likely AI-driven Direct traffic.
Step 3: Test AI tool referrer behavior yourself
Open ChatGPT (free and paid tiers), Gemini, and Claude. Ask each a question your content answers. Click the citation. Open DevTools → Network → look at the request headers. Note the Referer value (or absence). This tells you exactly which AI tools preserve referrer for your site.
Step 4: Tag manually where you can
For high-value content you submit to AI training datasets or publish on platforms AI assistants are likely to cite, append a UTM parameter:
?utm_source=ai_citation&utm_medium=referral&utm_campaign=ai_auditThis ensures even free-tier referrer-stripped clicks have attribution.
What to do with the data once you have it
If AI Assistant becomes a meaningful channel (>5% of traffic), three actions:
- Optimize for AI citation — clear definitions, structured data, FAQ schema, citable single-sentence claims
- Track conversions by AI channel — mark a key event, segment by AI Assistant in Explorations
- Set up a custom dimension for "AI source" — capture
chatgpt,gemini,claudeseparately so you can see which assistant drives which conversions
The bigger picture
AI Assistant traffic is small but high-intent — readers who clicked an AI citation already vetted your content as an answer. Conversion rates on this channel are typically 2–4× organic. Even if it's 5% of your traffic, it can be 15% of your conversions. Getting the attribution right matters for budget allocation and content strategy.
Audit your channel attribution and AI traffic capture
The new AI Assistant channel is one of several attribution changes that require updated audit logic. Run a free Snifflytics audit to verify your channel grouping, referrer handling, and UTM tagging are all configured to surface AI traffic correctly — before you start making content decisions based on the new data.