Google Analytics shipped a new default channel group this month: "AI Assistant." No opt-in, no config panel, no migration script to run. If a session's referrer matches a recognized AI chatbot domain, GA4 now tags the medium as ai-assistant, drops the campaign name as (ai-assistant), and buckets the session under the AI Assistant channel instead of Referral.
Google named ChatGPT, Gemini, and Claude as examples in the changelog. The full referrer match list isn't published anywhere I can find, and the Default Channel Group definitions page (support.google.com/analytics/answer/9164320) hasn't been updated yet to include AI Assistant in its channel table. So the exact matching logic is a black box right now. Worth flagging to any client who asks "how does it know."
Why this matters more than it looks
For the last year, every GA4 property I've touched has had some version of a manual regex rule stuffed into a custom channel group to catch chatgpt.com, gemini.google.com, or claude.ai referrers and pull them out of the generic Referral bucket. That was the only way to see this traffic distinctly. It worked, but it was fragile: miss a new chatbot domain, and it silently falls back into Referral or Direct. Same basic failure mode as cross-domain tracking turning real sessions into Direct: the report still renders, but the channel label stops telling the truth.
Google just did that work for you, system-wide, retroactively.
That's the part that needs attention. This isn't a new report you can ignore until you're ready. It changes the Default Channel Group logic on every GA4 property, which means:
- Historical-looking reports (last 30/90 days pulled today) will show different Referral totals than they did before the update, because some of that traffic just got reclassified as AI Assistant.
- Any conversion path or assisted-conversion report that references Default Channel Group will shift. If those numbers already feel unstable, run the same checks I use for diagnosing bad GA4 attribution before you explain the change to a client.
- Media-mix or MMM feeds pulling GA4 channel data as an input will see a new category appear from nowhere, which breaks any hardcoded channel list in a downstream model.
- Custom channel groups you built to catch this traffic manually will now double-count or conflict with Google's native version, depending on rule order.
None of this requires an admin to click anything. That's why it belongs on a QA checklist, not a feature announcement to skim past.
The gap Google didn't close
AI assistant traffic that arrives without a referrer header still lands in Direct. This happens with in-app browsers (ChatGPT's mobile app, for instance), copy-pasted links from a chat session, or any client that strips referrer info before the request hits your server. So "AI Assistant" channel volume is a floor, not a ceiling, on actual AI-driven traffic. Anyone reporting "AI referral traffic is only 2% of sessions" off this channel alone is underselling it. Direct still needs a manual audit for chatbot fingerprints (landing page patterns, UTM-less sessions with unusual engagement time, entry pages that only make sense as an AI citation destination).
What to do this week
1. Pull a before/after Referral comparison. Go into GA4, run Traffic Acquisition by Default Channel Group, and compare the last 7 days pre-update against the same window this week. If Referral dropped and AI Assistant appeared with volume, you've confirmed the reclassification hit your property. Screenshot it before your client asks why last month's Referral numbers "changed."
2. Audit custom channel group rules. If you built a manual rule to catch AI chatbot referrers (most consultancies did, using regex on source matching chatgpt|gemini|claude|perplexity), check whether it's now conflicting with the native AI Assistant channel. Custom channel groups evaluate before default groups in GA4's processing order, so your old rule may still be firing and masking the native categorization, or double-bucketing sessions depending on how you scoped it. Decide whether to retire the custom rule or keep it as a redundant check while Google's referrer list is still undocumented.
3. Flag it in any dashboard or MMM feed that hardcodes channel lists. If you've got a Looker Studio dashboard, a BigQuery export query, or a media-mix model that lists channels explicitly (Organic Search, Paid Search, Direct, Referral, Email, etc.), add AI Assistant to that list now. Otherwise it either gets dropped silently from aggregate rollups or gets bucketed into "Other," which understates it. This is also the kind of channel taxonomy drift that makes marketing and finance revenue reports stop matching.
4. Set up a Direct-traffic audit for AI fingerprints. Since no-referrer AI traffic still lands in Direct, build a segment for sessions with no source/medium that land on high-value pages with engagement time above your site average and no scroll-to-exit pattern typical of accidental Direct hits. It won't be perfect attribution, but it gives you a directional estimate of the gap the native channel misses.
5. Watch the Default Channel Group definitions page for the actual matching rules. Google Analytics Help hasn't published the channel table update yet. Once it lands, the referrer domain list will tell you exactly what's being caught and what isn't, which matters for anyone running attribution on emerging AI tools beyond the three named examples.
This is one of the few GA4 changes this year that closes a real measurement gap instead of adding another config surface to maintain. The retroactive part is what catches people off guard. Check your numbers before a client does.
Sources
- https://support.google.com/analytics/answer/9164320?hl=en
- https://searchengineland.com/google-analytics-ai-assistant-477544
- https://www.searchenginejournal.com/google-analytics-adds-ai-assistant-as-default-channel-group/574974/
- https://martech.org/ga4-now-tracks-ai-chatbot-traffic-automatically/