How to Measure AI Search Traffic: GA4, Search Console and ChatGPT Referrals
How to measure traffic and visibility from ChatGPT, Gemini, Copilot and Google's AI Overviews using GA4, Search Console, UTM tags and server logs.
Quick answer
Measure AI search with four sources. Search Console's generative AI reports show impressions in Google's AI Overviews and AI Mode. GA4's AI Assistant channel groups visits from ChatGPT, Gemini, Copilot, DeepSeek, Grok and similar assistants (Google's AI features stay in Organic Search). UTM and referrer data, including the `utm_source=chatgpt.com` tag ChatGPT adds to links, fills gaps. Server logs and monthly answer checks show crawler activity and how assistants describe you. Report them together as monthly trends by landing page and conversion, not as one vanity number.
What each tool can and cannot see
| Source | Shows | Does not show |
|---|---|---|
| Search Console: generative AI reports | Impressions of your pages in AI Overviews and AI Mode by page, country, device and date | Other assistants; the answer text; brand mentions without a link |
| Search Console: Performance (Web) | Clicks and impressions for all Google Search, including AI features | A clean split of clicks by AI feature |
| GA4: AI Assistant channel | Sessions and conversions referred by ChatGPT, Gemini, Copilot, DeepSeek, Grok and similar | Google AI Overviews and AI Mode (counted as Organic Search); visits with no referrer |
| UTM parameters | ChatGPT links tagged utm_source=chatgpt.com; links you control in AI apps or feeds | Untagged links from other assistants |
| Bing Webmaster Tools | Bing and Copilot search performance | Non-Microsoft assistants |
| Server or CDN logs | Which AI crawlers and fetchers request which pages | Whether a page was cited |
| Manual or tool-based answer checks | How assistants describe you for chosen questions | Representative coverage; results vary by user and time |
Set up GA4 for AI traffic
Since May 2026, GA4's default channel group includes an AI Assistant channel. Google defines it as the channel through which users arrive from sources such as ChatGPT, Gemini, DeepSeek, Copilot or Grok; sessions match when the medium is `ai-assistant` or the referrer is on Google's list of AI assistant sources. It explicitly excludes Google's AI Overviews and AI Mode, which remain in Organic Search. No configuration is needed for the default group.
If you built a custom channel group with regex rules before this change, check the order of rules. GA4 evaluates channels top to bottom; a custom AI rule placed below Referral never matches. Many teams can now retire their custom rule and use the default channel, keeping a custom group only if they need finer splits (for example ChatGPT separately from Perplexity).
- Open Reports › Acquisition › Traffic acquisition and confirm the AI Assistant row appears
- Add session source as a secondary dimension to split assistants
- Mark key events (lead form, purchase, demo request) so the channel shows conversions
- Create an exploration of AI Assistant sessions by landing page
- Annotate the date your AI-related content changes went live
Use Search Console's generative AI reports
Google introduced Search Generative AI performance reports in Search Console on 3 June 2026, first for a subset of properties and then more widely. They give a dedicated view of impressions for your URLs in generative AI features on Search (AI Overviews and AI Mode) and in Discover, broken down by page, country, device and date.
Use them to answer two questions: which pages Google's AI features draw on, and whether that is growing. Compare the AI impression trend with the page's overall Search performance. A page with rising AI impressions and falling clicks may be answering the question well enough inside the result; improve the reasons to click (tools, detail, pricing, next steps) rather than hiding the content.
Worth noting
Google's guide states that no third-party tool has access to its internal ranking or AI systems. Search Console is the only first-party source for Google AI feature visibility.
Tag, log and fill the gaps
ChatGPT automatically adds `utm_source=chatgpt.com` to links in search answers, according to OpenAI's publisher FAQ, so those visits are identifiable even when the referrer is missing. Other assistants vary: some pass a referrer from the web version but not from mobile or desktop apps, which pushes visits into Direct. Look for unusual Direct sessions landing deep in the site (on a comparison guide or a specific product) as a signal of untagged assistant traffic.
Where you control the link (product feeds submitted to AI shopping channels, links in your own AI app or MCP server), add consistent UTM parameters. Do not add UTMs to internal links on your own site; that overwrites the original source.
Server or CDN logs complete the picture by showing which AI crawlers visit and what they request. A sudden drop in OAI-SearchBot or Claude-SearchBot requests usually means a robots.txt or firewall change, not a ranking change. See AI crawlers and robots.txt for what each bot does.
Track how assistants describe you
Traffic numbers miss answers that mention you without a click. Keep a short list of 10 to 30 questions that matter commercially ("best [service] for [industry]", "[your brand] vs [competitor]", "how much does [service] cost") and check them monthly in ChatGPT, Gemini, Copilot, Perplexity and Google AI Mode. Record the date, whether you are mentioned, whether you are linked, and whether the description is accurate.
Monitoring tools can automate this at scale. They sample prompts and are useful for trends, but answers vary by user, location and wording, so treat their share-of-voice figures as estimates, not measurements.
A monthly AI search report
Keep it short and comparable month to month.
| Metric | Source | Why it matters |
|---|---|---|
| AI feature impressions (Google) | Search Console generative AI report | Visibility trend in Google's AI answers |
| Top pages by AI impressions | Search Console | Which content Google's AI draws on |
| AI Assistant sessions and conversions | GA4 | Whether assistant visits create business value |
| Top AI landing pages | GA4 exploration | Where to improve next steps and conversion |
| AI crawler activity | Server or CDN logs | Early warning for access problems |
| Answer check results | Manual log or monitoring tool | Accuracy of how you are described |
Want AI channels in your reporting?
ZSpace Labs sets up GA4, Search Console and log reporting for AI search and connects it to conversion analysis. See our CRO and analytics audit.
Interpreting the numbers
AI assistant traffic is still small for most sites compared with organic search, but it often converts differently because visitors arrive after an assistant has narrowed their options. Compare conversion rate and lead quality by channel rather than session counts. Watch for these patterns:
- AI impressions up, clicks flat: the answer satisfies the question; give people a reason to visit (tools, pricing, detail)
- Assistant visits land on old pages: update or redirect them; assistants may cite outdated URLs
- Inaccurate descriptions in answer checks: fix inconsistent facts on your site and profiles
- Crawler requests drop suddenly: check robots.txt, CDN bot settings and server errors
Conclusion
AI search is measurable enough to manage, as long as you combine sources. Use Search Console for Google's AI features, GA4's AI Assistant channel for other assistants, UTMs and logs to fill gaps, and regular answer checks for accuracy. Report trends by page and conversion. For the work that moves these numbers, see how to make your website discoverable in AI search, and for wider measurement design, our guide to ecommerce attribution.
Common questions.
Yes. GA4's default channel group now includes an AI Assistant channel for visits from sources such as ChatGPT, Gemini, DeepSeek, Copilot and Grok. ChatGPT also adds utm_source=chatgpt.com to links in its answers, which helps attribution.