Foundation · 2 min
What you need to know
Can you count every mention?
No complete cross-service count is available. You can record answers to a fixed set of questions at regular intervals, noting the service, date, wording, and cited pages. Results can vary, so treat the sample as directional.
Which measures matter?
Track relevant referrals, useful visits, orders, and corrected catalog errors alongside search visibility. A rise in mentions with no qualified visits is only one signal.
How can you judge whether a change helped?
Record the page and date of the change, watch the same queries and traffic over time, and note other changes such as seasonal demand or a promotion. Avoid claiming that one edit caused a traffic shift without stronger evidence.
What can you put in a monthly report?
Record referral visits that identify a source, the landing pages those visitors use and whether they complete useful actions. Also run the same small set of shopping questions and record whether your business appears and whether the facts are correct.
Do not turn a handful of prompts into a market-share percentage. Instead, compare like-for-like checks over time and investigate inaccurate answers that could harm a buyer.
What can measurement not prove?
A referral visit can show that someone arrived from an identifiable source, but it may not reveal the exact answer or prompt that influenced them. A manual prompt check can show an answer at one moment, but results may vary later.
Report each measure with its limitation. Avoid claiming that one content edit caused a sale unless the evidence supports that conclusion.
Foundation · 2 min
Frequently asked questions
Can analytics show every AI mention?
No. Some services do not send identifiable referral data, and answers can vary. Report the limits of the measurement.
What counts as improvement?
Use a defined outcome, such as more qualified visits, accurate product mentions, or completed orders from an identifiable source and compare consistent periods.