AI search analytics: measure citations without inventing a dashboard

Log the referrals you can actually see

Some chat tools send a referrer, and some do not. Record the ones that arrive, but do not treat an empty referrer report as proof you are absent.

Watch branded queries and direct visits when a major guide is being quoted. The visit may come later, as a search for the name, rather than as a click from the summary.

Keep a citation sheet for priority queries

For a short list of queries, note the date, the engine, whether you were named, and which URL was cited. Ten queries checked monthly beat a tool that scores ‘AI visibility’ with a made-up index.

Tie each query to the URL that should own it. If the wrong URL on your own site is cited, you have a cannibalisation problem inside the summary.

• A fixed query list.
• The URL you expected to see.
• A note when the cited fact is wrong.

Act on errors and absences differently

An incorrect citation means fix the source trail. An absence on a query you rank for means tighten the passage. An absence on a query you do not deserve means leave it.

Review the sheet when you refresh a cornerstone. Measurement that never changes the next edit is only a scrapbook.

Is there a reliable AI rank tracker?

Treat automated scores as hints. Interfaces change, locations differ, and many answers are personalised. A logged manual check of priority queries is more honest.

What is a good AI referral number?

There is no benchmark that matters more than your own trend. Pair it with whether the right URL is cited for the queries you chose.