# How to track AI referral traffic in GA4

AI visibility is easy to celebrate and hard to prove. The proof lives in your analytics: how many humans clicked through from an AI answer, and what they did next. This guide explains how to read that in Google Analytics 4 — and why the default reports undercount it.

## Why default GA4 misses AI traffic

Out of the box, GA4 groups traffic by channel. AI platforms usually land in **Referral** or **Organic Social** depending on the model and the URL parameters. Some AI apps send no referrer at all, so sessions appear as Direct.

Two consequences:

- An AI referral can look like a direct visit, inflating Direct and hiding the AI channel.
- Aggregated channel groups bury the source, so you cannot tell ChatGPT from a news site.

The fix is to look at **session source** (`sessionSource`) directly, not at the default channel grouping.

## Step 1: Confirm the property and access

You need the GA4 property ID (Admin → Property settings) and read access. If you want to automate the pull, create a service account in Google Cloud, grant it Viewer access on the property, and download its JSON key.

## Step 2: Query by session source

In GA4, open **Reports → Acquisition → Traffic acquisition**, switch the dimension to **Session source**, and look for these hosts:

| Platform | Typical session source |
|---|---|
| ChatGPT | `chatgpt.com` |
| Perplexity | `perplexity.ai` |
| Gemini | `gemini.google.com` |
| Claude | `claude.ai` |
| Copilot | `copilot.microsoft.com` |
| DeepSeek | `deepseek.com` |
| Grok | `grok.com` |

Anything not on this list but ending in one of these domains is still AI traffic. Match on the tail of the host, not an exact string.

## Step 3: Add conversion and revenue metrics

Sessions alone are vanity. Add **active users**, **conversions**, and **total revenue** with the same session source dimension. That answers the only question leadership asks: *does AI traffic turn into money?*

## Step 4: Segment landing pages

Repeat the query with `landingPage` as a second dimension. This tells you which page AI is sending people to. High-traffic landing pages from AI are your citation winners — protect and expand them. Pages AI never sends traffic to are candidates for an answer-shaped rewrite.

## Step 5: Automate the pull (optional)

If you want a weekly number instead of a manual report, use the GA4 Data API with a service account and run a report on `sessionSource` with `sessions`, `activeUsers` and `totalRevenue`. Cache the result; the API has quotas.

Bidvix's AI traffic tab does exactly this: connect a property, and it filters the 18 most common AI sources, then shows sessions, users, revenue and top landing pages.

## Common pitfalls

1. **Counting bots.** AI crawlers fetch pages but do not appear as sessions in GA4 unless they execute JavaScript. Revenue and users are the honest signals.
2. **Missing direct traffic.** Some AI apps strip the referrer. Treat a sudden Direct spike alongside a rise in AI citations as a probable AI effect, and confirm with a UTM-tagged link where possible.
3. **Comparing periods incorrectly.** AI traffic is lumpy. Compare four-week windows, not week over week.
4. **Confusing citations with clicks.** A citation in an answer may never be clicked. Citation rate and referral sessions are different metrics; track both.

## What good looks like

- AI referral sessions trend up while your mention rate also rises.
- Revenue from AI sources grows faster than session count (higher intent).
- Top AI landing pages match the pages you deliberately made answer-shaped.

If sessions rise but revenue does not, your landing pages convert poorly for AI-referred intent. That is a content problem, not a tracking problem.

## Related reading

- [What is AI share of voice?](/answers/what-is-ai-share-of-voice)
- [How do I know if AI cites my website?](/answers/how-do-i-know-if-ai-cites-my-website)
- [What is generative engine optimization?](/guides/what-is-generative-engine-optimization)