Measure the traffic AI engines actually send you
Your analytics show a growing trickle from chatgpt.com, perplexity.ai, copilot.microsoft.com. The question worth answering: which answers sent it, and could you get more?
What you can measure directly
Referrer data from AI engines is partial but useful:
- Perplexity and ChatGPT send referrers when users click citation links. Tag them as a channel group; watch the trend, not the daily noise.
- Landing pages tell the story. Which of your URLs receive AI referrals is ground truth for which pages get cited — cross-check against
sources[]from your monitor runs. - Google AI Overview doesn’t send a distinct referrer. Clicks arrive as regular Google organic. The proxy: impressions/clicks on queries where an AIO was shown — which you know because you monitor which of your keywords trigger one.
What you can’t measure
Most AI influence is dark: the user reads an answer, doesn’t click, and later searches your brand directly or types your URL. Expect brand-search volume and direct traffic to rise ahead of measurable referrals. That lag is normal — it means citations are working.
Closing the loop
The useful dataset joins two sides:
- What the answers say — your monitored prompt set: mentions, citations, position. (That’s the monitor API.)
- What users did — referrals per cited URL, brand-search trend, assisted conversions.
When a prompt’s citation share goes up and the cited page’s referrals follow, you’ve found the lever. When citations rise but traffic doesn’t, the answer is satisfying the intent in-line — optimize for the mention, not the click.
AI referral measurement is messy by nature. The teams that win treat the answer itself as the channel — and traffic as the lagging indicator.