AI Referral Share Reader

A free tool for anyone asking whether AI search sends them customers. Paste a Google Analytics 4 Traffic acquisition export, by session source / medium, and the reader separates the visits and key events sent by ChatGPT, Perplexity, Gemini, Copilot, Claude, DeepSeek and Grok. It shows their share of the site, and whether those visits really convert differently from the rest of the site or the data is too thin to tell. It runs in your browser; nothing you paste is sent anywhere.

How it recognises an assistant

A row counts as an assistant when its source is a web address the assistant's maker publishes. ChatGPT is chatgpt.com: OpenAI's Publishers and Developers FAQ says ChatGPT adds utm_source=chatgpt.com to referral links. Gemini is gemini.google.com, from Google's Gemini Apps Help. Copilot is copilot.com, Claude is claude.ai, DeepSeek is chat.deepseek.com and Grok is grok.com, each from its maker's own pages. Perplexity is perplexity.ai, its own domain. A row whose medium is ai-assistant, which GA4 sets for AI assistant visits, counts too. You can add other assistants by their address.

Is the difference real?

A source with a handful of sessions can show a striking conversion rate by chance. So the reader compares the assistants with the rest of the site as a ratio with an exact 95% interval: sessions with a key event when the export has Session key event rate, otherwise key events per session. It says higher only when the whole interval is above 1, lower only when it is below 1, about the same when it sits inside 0.9 to 1.1, and otherwise that there isn't enough data yet. For single assistants, the 5% is split across the assistants compared. The test, the 95% and the 10% band are AAJ's choices.

What it can't see

• Google's AI Assistant channel excludes AI Overviews and AI Mode, so those clicks are counted as Google search.
• Visits from apps that pass no referrer usually arrive as direct traffic.
• The comparison is about visits, not people or deals. Without Session key event rate, one visit can trigger several key events, which makes the interval somewhat too narrow.

The share it reports is a floor, not the full AI effect.

Why I built it

Teams see chatgpt.com in their referrals, notice a high conversion rate on a few dozen visits, and call it a channel. Sometimes it is. Usually it's too early to say. I'd rather the number came with an honest "can't tell yet" than a headline that doesn't hold up next quarter.

The same engine runs as an agent skill, ai-referral-share. To tag the links you control, use the UTM Builder and Naming Checker. To see whether AI answers name you, use the AI Visibility Significance Test.

Run this in your agent

The free AAJ agent skill AI Referral Share does this job inside a coding agent, on your own data. Install it with npx skills add sarojkjha/aaj-marketing-skills --skill ai-referral-share.

Using Claude or ChatGPT instead? Add the AAJ engines as a custom connector, with no install: https://aaj-engines-mcp.sarojkjha10.workers.dev/mcp