On 13 May 2026, Google Analytics introduced a dedicated AI Assistant channel for traffic arriving from recognised assistants. Instead of leaving every click from ChatGPT, Gemini, Copilot or another supported source mixed inside Referral, GA4 can now assign the medium “ai-assistant” and the campaign “(ai-assistant)”. For Indian businesses investing in AI-search visibility, this creates a clearer way to measure whether those citations and recommendations produce useful website visits.
The new channel is not a complete “AI visibility score”. It measures identifiable clicks that reach the website with a recognised referrer; it cannot count an answer that mentions the brand without a click, and some apps or privacy settings can remove referral information. Google also classifies traffic from its own AI Overviews and AI Mode under Organic Search rather than AI Assistant. This guide explains how to build an honest dashboard from acquisition through qualified lead and revenue.
1. Confirm that the new channel is present
Open Reports, then Acquisition, then Traffic acquisition. Change the primary dimension to Session default channel group and look for “AI Assistant”. A property may show no row until it receives matching traffic. Use a date range after the May 2026 release and compare with the source and medium dimensions; older traffic may have been classified differently and should not be silently rewritten into a new trend line.
2. Understand GA4’s automatic classification
Google’s default channel documentation says recognised AI referrals receive the medium “ai-assistant” and campaign “(ai-assistant)”. The channel can include sources such as ChatGPT, Gemini, DeepSeek, Copilot and Grok. The recognised list and rules can evolve. Do not hard-code a permanent platform list in management reports without checking current GA4 definitions.
3. Separate AI assistants from Google AI search
The default AI Assistant channel excludes Google AI Overviews and AI Mode; those visits are included in Organic Search. Therefore the AI Assistant row cannot answer “How much traffic came from every AI-generated answer?” Report it as identifiable external assistant referrals. Keep Organic Search visible beside it and explain the classification to stakeholders before comparing channels.
4. Use Traffic acquisition for sessions
Traffic acquisition is session-scoped and shows how visits were acquired during the selected period. Add Session source beside Session default channel group to see which recognised assistants are sending sessions. Compare users, sessions, engaged sessions, engagement rate, average engagement time and key events. Avoid declaring a winner from five or ten visits; small samples are naturally volatile.
5. Use User acquisition for first discovery
User acquisition is user-scoped and answers how new users first arrived. An AI assistant may introduce a customer, while a later direct or branded-search visit completes the enquiry. Compare First user default channel group with Session default channel group rather than mixing the two. The same person can legitimately appear under different channels at different scopes.
6. Break performance down by landing page
Add Landing page plus query string as a secondary dimension or use the Landing page report with an AI Assistant comparison. Identify which guides, product pages, calculators, service pages and location pages attract cited clicks. Check whether the visitor sees a clear answer, credible proof and an obvious next action without having to repeat the research.
7. Mark only meaningful actions as key events
GA4 key events should represent actions important to the business: a qualified form submission, booked consultation, completed purchase, phone click with appropriate follow-up or another defined milestone. Scroll depth and page views are useful diagnostics but weak sales outcomes. Test each event, prevent duplicate firing and document the business meaning before using it in an AI-traffic report.
8. Calculate a simple AI referral funnel
Build one monthly table that progresses from traffic to value. Keep the definitions consistent across channels.
- AI Assistant sessions: sessions classified in the default AI Assistant channel.
- Engaged-session rate: engaged AI sessions divided by AI sessions.
- Key-event rate: sessions with the selected business action divided by AI sessions.
- Qualified-lead rate: CRM-qualified leads divided by identifiable AI referral enquiries.
- Customer value: realised contribution or revenue from matched customers, shown with attribution limitations.
9. Compare source quality—not only volume
One assistant may send fewer visits but better-qualified buyers because the user has already compared options in a long conversation. Review landing page, geography, device, engagement and downstream outcome by source. Do not publish a source leaderboard when privacy thresholds, low volume or inconsistent referrers make the differences unstable.
10. Know why some AI traffic appears as Direct
A user may copy a URL, open it from an app that removes the referrer, move between devices or use a privacy setting that limits attribution. Those visits can appear as Direct or another channel. GA4 cannot reconstruct context it never received. Treat measured AI Assistant traffic as a visible subset, not an exact count of every AI-influenced visit.
11. Do not use UTMs to overwrite organic referrals
You normally cannot control the links that an independent AI assistant cites. Do not add campaign parameters after the fact or rewrite genuine referral data to make a dashboard cleaner. Use standard UTMs only on links the business intentionally distributes—such as paid partnerships, email, creator campaigns or owned chatbot flows—and maintain a documented naming convention.
12. Connect website enquiries to the CRM
Store the first landing page, source, medium, campaign, enquiry time and consented contact identifier with the lead where appropriate. Add qualification stage, opportunity value and closed outcome later. Keep marketing attribution fields separate from personal notes. A channel that produces fewer forms but a higher proportion of sales-qualified opportunities may deserve more content investment.
13. Review assisted journeys
The Advertising workspace and attribution-path reports can help show touchpoints that occur before key events. Availability and report behaviour can vary by property and feature rollout. Look for patterns such as AI Assistant to Direct, AI Assistant to Organic Search or AI Assistant to Paid Search. Present these as observed paths, not proof that one touchpoint caused the sale.
14. Create a reusable GA4 exploration
Build a free-form exploration with Session default channel group, Session source, Landing page, Device category and Country or region. Use metrics for sessions, engaged sessions, key events and revenue where implemented. Apply an AI Assistant filter and save the exploration with the date and definition. Duplicate it for Organic Search so the team can compare behaviour without changing the base report every week.
15. Add a content-quality layer
For each landing page receiving AI referrals, record its topic, last factual review, author or reviewer, proof sources, conversion path and current offer. A spike to an outdated article can create traffic while damaging trust. Update facts, pricing and availability without erasing useful publication history, and keep structured headings that make the page easy for people to verify.
16. Report citations and traffic separately
An external monitoring tool may estimate brand mentions or citations, while GA4 records visits that reach the site. These are different datasets. A brand can be mentioned frequently but receive few clicks because the answer satisfies the user, or receive valuable clicks from a small number of high-intent citations. Never divide GA4 conversions by an unrelated estimated mention count without explaining the methodology.
17. Use a 30-day measurement routine
Week one: verify GA4 events, consent behaviour and CRM fields. Week two: baseline AI Assistant and Organic Search by landing page. Week three: improve two high-potential pages with clearer proof and next actions. Week four: compare engaged sessions, qualified enquiries and outcomes with the baseline. Record content changes so the team does not confuse seasonality with an optimisation effect.
18. Apply a scale-or-investigate rule
Scale content around a topic when identifiable AI referrals repeatedly reach the right page and produce useful actions. Investigate when traffic lands on irrelevant URLs, engagement is weak or CRM quality falls. Keep publishing for customer usefulness even when volume is small; the correct early objective is clean evidence and a repeatable learning loop, not a vanity traffic target.
Key takeaways
Put the framework into action
- GA4 now automatically classifies recognised external assistant referrals in a dedicated AI Assistant channel with the “ai-assistant” medium.
- Google AI Overviews and AI Mode remain in Organic Search, while stripped referrers and no-click mentions are not captured as AI Assistant visits.
- Measure landing-page quality, key events, qualified leads and customer outcomes instead of reporting AI traffic volume alone.
Research sources
Sources & further reading
This article provides general business education, not legal, tax or financial advice. Adapt the framework to your market and consult qualified professionals where required.