AI referral traffic attribution: a GTM and GA4 setup that holds up
Set up reliable attribution for AI-referred visits using GTM and GA4, learn what the data cannot tell you, and see where a dedicated visitor analytics script fills the gaps.
Quick verdict
AI traffic hides in analytics. Some arrives with a clear referrer, some arrives as direct, and some is not clicked at all but still influences a purchase. You can capture a good share with a few GA4 and GTM settings, and it is worth doing this week. For conversions tied to specific AI engines, a purpose-built script is easier to keep honest. Promptwatch offers visitor analytics for that, through a lightweight script, a GTM template, or the WordPress plugin.
What you are trying to measure
| Question | Hard to answer? |
|---|---|
| How many sessions came from AI engines? | Medium, referrers help |
| Which engine sent them? | Medium |
| Did they convert? | Easy once sessions are labeled |
| How many were influenced but arrived as direct? | Hard, not fully recoverable |
| Which page was cited? | Needs citation data, not analytics |
Be clear with stakeholders: attribution will undercount. Some AI answers produce a branded search a day later, and no setup recovers that.
Step 1: list the referrers
AI products send referrers with names you can match. Start with the engines you track: ChatGPT, Perplexity, Gemini, Claude, Copilot, DeepSeek, Kimi, and Mistral. Use the domains you actually see in your referral report, because domains change. Check your reports, rather than trusting a list from a blog post, including this one.
Step 2: create a GA4 channel group
In GA4, build a custom channel group with a channel named "AI search" and a rule matching source against the domains from step 1. Order matters: place it above the default Referral channel, or sessions land in Referral first.
Test it by checking recent sessions: do the visits you expect appear under the new channel?
Step 3: tag your own links
When you can control the link, use UTM parameters. Example: a link in a newsletter. You cannot control the link an AI engine shows, so UTMs only help in channels you own, such as your own campaigns. Do not add parameters to pages you want cited in a way that creates duplicate URLs.
Step 4: use GTM for event consistency
Google Tag Manager lets you fire conversion events consistently. Make sure these fire correctly:
- Form submissions, with a clear event name.
- Signup completions.
- Purchases, if you sell online.
- Demo requests.
Mark the important ones as key events in GA4, so that AI search conversions can be reported by channel.
Step 5: check for leakage
Look at the Direct channel. A surge on deep pages with no referrer, for pages that tend to be cited, suggests AI visits are leaking into Direct. Some apps strip referrers. Document the gap and report the AI number as a floor.
Where a visitor analytics script helps
GA4 is general-purpose. A script designed for this job removes manual rule maintenance. Promptwatch's visitor analytics records AI-referred traffic and conversions from ChatGPT, Perplexity, Gemini, Claude, Copilot, DeepSeek, Kimi, Mistral, and more. Install with a lightweight script, the GTM template, or the WordPress plugin. It is available from Essential, with 200K visitor events, and scales to 1M on Professional and 10M on Business.
The deeper benefit is joining. 360 Insights combines crawler logs, citation clicks, and AI-referred visits per page. That lets you ask which cited pages produce visits and which are cited without a click.
Reading the join
| Pattern | Meaning | Action |
|---|---|---|
| Cited often, few visits | Answer satisfies without a click, or low source position | Add a distinctive angle |
| Few citations, steady visits | Existing visits from a few strong pages | Protect them, expand |
| Visits, no conversions | Landing mismatch | Align page with the question |
| Visits and conversions | Working page | Refresh on schedule |
Evidence from customers
Promptwatch's published case studies show how customers use it, and they are the customers' results, not guarantees. Crisp reported AI-referred traffic converting at twice the rate of traditional channels. TreeHouse, behind Pararius and Huurwoningen.nl, reported AI search becoming about 25% of organic conversions, with 99% of that from ChatGPT, after fixing crawler access and automating weekly reporting with the Promptwatch MCP. Neither number means your site will see the same. Both show why attribution is worth the setup.
Plan notes
Explore is free and has no visitor analytics we can confirm, so check before relying on it. Essential at $95 a month includes 200K visitor events. Professional at $245 includes 1M events and 25M crawler logs. Business at $579 includes 10M events. Agency plans include visitor analytics from $199 a month, with 500K events on Kick-off.
Other tools in the ranking handle attribution differently. Gauge's directory entry lists GA4 and Search Console connections, for example.
Who should pick which
| You are... | Recommendation |
|---|---|
| Starting from zero | GA4 channel group and key events this week |
| Needing engine-level conversions | Promptwatch visitor analytics |
| Reporting to finance | Report a floor, add the leakage note |
| An agency | Agency plans, which include visitor analytics |
| Running WordPress | The plugin, which supports visitor analytics and crawler logs |
Set up the GA4 channel today because it costs nothing. Add Promptwatch's visitor analytics when you need conversions by engine and want that data beside citations and bot logs in one view.