Trial and error taught me how to measure AI visibility properly

For six months I treated it as a counter: mentioned or not. The number went up and I had no idea what to do with it.

Then I added three things to my AI visibility dashboard: sentiment, the competitors named in the answer, and the sources the model cites. The work on my client's site changed after that.

• Sentiment surfaced the negative reviews the models were picking up in their answers. We handled them and replied publicly.

• The competitors turned out to be weaker than they looked in search. A small budget spent on page usability and on text put the site next to the biggest players.

• The sources showed where to go. A review in one place, a genuinely useful comment that collected upvotes in a thread, a few YouTube videos with the keywords in the title.

It turned out this is not a mention counter. It is a list of specific places you need to get into.

What are you using that I do not have?