Most sites that get traffic from ChatGPT, Gemini, Claude or Perplexity find it is oddly concentrated. A handful of pages, often not the ones you would guess, receive most AI referrals. Knowing which pages those are, and what they have in common, is one of the most practical things you can learn about AI search.
This guide shows you how to build a Google Analytics 4 exploration of AI referral landing pages, step by step, using the AI Assistant channel Google added in 2026 plus a source filter for older data. Then we cover how to read the results and turn them into content patterns you can replicate across your site.
If you have not set up AI tracking at all yet, start with our guide to tracking ChatGPT and AI traffic in GA4.
Key Takeaways
- GA4's landing page is the first page viewed in a session, which makes it the right dimension for analysing where AI visitors arrive.
- Since May 2026, GA4 assigns recognised AI assistant referrals to an "AI Assistant" channel; a source regex helps with older data and unlisted tools.
- A free-form exploration with landing page rows and engagement and key event metrics is enough for most sites.
- Tag each landing page by type, format and topic to spot patterns rather than isolated winners.
- Small numbers are normal; widen date ranges and avoid over-reading pages with only a few sessions.
What counts as an AI referral in GA4
Google's May 2026 release note explains that a visit receives the medium ai-assistant and the "AI Assistant" channel when its referrer matches a recognised AI assistant, with ChatGPT, Gemini and Claude given as examples. Google has not published the full list, and the note does not say whether older data is reclassified, so check your own reports.
Two groups are missing from any referrer-based report: AI visits that arrive without a referrer, which we explain in why AI traffic shows up as Direct, and clicks from Google's AI Overviews and AI Mode, which arrive as Google traffic. Keep that in mind when you present results: this is a view of identifiable AI referrals, not all AI influence.
Build the exploration step by step
Google's help page defines the landing page dimension in one line.
“The page path and query string associated with the first pageview in a session.”
— Google Analytics Help, Landing page report
Now build the report. Our general guide to GA4 explorations covers the interface if you are new to it.
- Open Explore and start a new Free form exploration. Name it "AI referral landing pages".
- Set the date range to at least the last 90 days. AI referrals are often small, and short ranges trigger data thresholds.
- Import dimensions: Landing page + query string, Session default channel group, Session source / medium and Device category.
- Import metrics: Sessions, Engaged sessions, Engagement rate, Average engagement time per session, Key events and Session key event rate. Add Total revenue for ecommerce.
- Rows: drag in Landing page + query string. Set "Show rows" to 100 or more.
- Values: add the metrics above. Sort by Sessions.
- Filter: Session default channel group exactly matches "AI Assistant". For older data or extra tools, use a second version of the tab with a filter on Session source matching a regex such as
chatgpt|openai|perplexity|gemini|claude|copilot. - Compare segments (optional): add a segment for AI sessions and one for Organic Search sessions to see the same pages side by side. Free-form explorations support up to four segments.
- Add a breakdown: put Session source in the columns to see which assistant sends traffic to each page.
- Save and share, then repeat monthly so you can track changes.
Here is the same setup as a quick reference.
| Setting | Value | Notes |
|---|---|---|
| Technique | Free form | Table visualisation |
| Rows | Landing page + query string | Keeps UTM evidence visible |
| Columns (optional) | Session source | Shows which assistant sent each visit |
| Values | Sessions, Engaged sessions, Engagement rate, Key events, Session key event rate | Add Total revenue for ecommerce |
| Filter A | Session default channel group = AI Assistant | For data since the channel launched |
| Filter B | Session source matches regex of AI domains | For older data and unlisted tools |
| Segments (optional) | AI sessions vs Organic Search sessions | Up to four segments can be compared |
Google's help notes that filter clauses in free-form explorations use AND logic, so build an OR condition with a regex rather than two separate filters. Its article on exploration differences also says data may be withheld when user counts are low, and that exploration date ranges are limited by your property's data retention setting.
Turning the list into patterns
A ranked list of pages is a start. The insight comes from tagging each page and looking for common traits. Export the top 30 to 50 AI landing pages and add columns for these attributes.
| Attribute | Examples of values | Why it matters |
|---|---|---|
| Page type | Guide, comparison, product, pricing, glossary, tool, case study | Shows which formats AI tools link to |
| Primary intent | Learn, compare, buy, troubleshoot | Shows where in the journey AI sends people |
| Unique asset | Original data, calculator, template, checklist, none | Original material is harder for AI to replace |
| Answer format | Definition first, steps, table, list | Links back to passage-level structure |
| Freshness | Last updated date | Recency may matter for some topics |
| Topic cluster | Your own content clusters | Shows topical strengths |
| Also gets organic traffic? | Yes or no, from the Organic Search segment | Separates AI-specific winners from all-round winners |
Then look for repeated combinations. A typical finding might be that comparison pages and pages with original numbers attract AI referrals out of proportion to their organic traffic. Our upcoming guide to statistics pages for AI citations looks at that format in detail, and our post on text fragment links shows how to see the exact passages Google features highlight.
Questions to ask of the data
- Which pages get AI referrals but little organic traffic? They may answer questions people now ask AI rather than Google.
- Which pages get both? They are proven assets; keep them updated.
- Which pages have high AI sessions but low engagement? The visitor may have expected something different. Check the opening paragraph and title.
- Which landing pages lead to key events? Our guide to GA4 key events shows how to set these up correctly.
- Are query strings revealing
utm_source=chatgpt.com? OpenAI says ChatGPT adds it, which confirms the source even when the referrer is weak.
Acting on what you find
- Replicate winning formats on related topics where you have no page yet.
- Improve the top AI landing pages for the visitor who arrives mid-research: clear summaries, next-step links and a visible call to action. Our landing page SEO guide covers this balance.
- Add internal links from AI landing pages to commercial pages.
- Track the same pages' engagement and conversions over time, and compare with other channels as described in testing AI traffic quality.
- Put the exploration's key numbers into an AI search dashboard in Looker Studio.
Common mistakes
- Using the Page path dimension. It counts all pageviews, not where sessions started. Use Landing page.
- Filtering on referral medium only. AI traffic may now carry the
ai-assistantmedium, so a "referral" filter can miss it. - Reading tiny numbers. A page with three AI sessions is not a pattern.
- Ignoring query strings. Without the "+ query string" dimension, UTM evidence disappears.
- Forgetting hidden AI visits. Referrer-based data is a lower bound; see our guide to referral traffic in GA4.
- Copying page length. Look at structure and unique value, not word count.
Related Guides
- Is AI Referral Traffic Higher Quality? How to Test It on Your Own Site
- Building an AI Search Dashboard in Looker Studio (Now Data Studio)
- How to Choose an AI Visibility Tracking Tool: Questions to Ask
Frequently Asked Questions
How do I see which pages get ChatGPT traffic in GA4?
Create a free-form exploration with Landing page + query string as rows, filter Session default channel group to AI Assistant or Session source to chatgpt, and add sessions and key event metrics.
Why are my AI landing page numbers so small?
AI referrals are still a small channel for many sites, and some AI visits arrive without a referrer. Use longer date ranges and focus on patterns across many pages.
Does the AI Assistant channel include Google AI Overviews?
Google's release note describes the channel for referrers from AI assistants such as ChatGPT, Gemini and Claude. Clicks from AI Overviews and AI Mode come from Google Search and are reported in Search Console.
Should I use Landing page or Page path?
Use Landing page. It records the first page in each session, which is where the AI visitor arrived.
How often should I update this analysis?
Monthly for the report, and quarterly for the pattern review, since AI tools and your content both change.
Conclusion: find the pattern, then build more of it
A simple GA4 exploration can reveal which of your pages AI assistants send visitors to. Tagging those pages by type, format and unique value turns that list into a content plan. If you want help creating more pages that earn AI referrals, our content writing service can build them with you. Get a free quote to start.
References
- Google Analytics Help: Landing page report
- Google Analytics Help: What’s new in Google Analytics
- Google Analytics Help: Free-form exploration
- Google Analytics Help: Data differences between reports and explorations
- Google Analytics Help: About data thresholds
- OpenAI Help Center: Publishers and developers FAQ
