You may be making budget decisions on the wrong number
AI visibility has moved from a curiosity to a budget line. Reports now go to leadership, and the numbers in them decide what gets funded next quarter. That makes it worth asking what those numbers actually count, because a figure can rise steadily while buyers see no more of you than before.
Two ways a number flatters you
The first is counting the wrong event. Many dashboards report crawler visits, pages indexed and answer citations with similar words, and only the last one means an engine showed you to someone. Felix covers the difference in his piece on what dashboards count.
The second is counting the wrong prompts. If most of the prompts tracked already contain your brand name, you will appear in most of the answers, and the score will look strong. Héloïse explains how to separate them in her piece on branded and unbranded visibility.
Either one on its own can produce a healthy-looking report. Together they can make a brand that no new buyer ever hears about look like a category leader.
What this does to a budget
A number that rises regardless of what you do cannot tell you which work paid off. So you end up funding whatever was done most recently, or whatever the vendor recommends, and the report agrees with every choice. The money is not wasted on its own terms. It is simply unaccountable.
Four questions before the next budget line
Ask whoever produces the report these four things. They should be easy to answer.
| Question | A good answer |
|---|---|
| What event does this number count? | An engine naming or citing us in an answer |
| How many of the tracked prompts contain our name? | Reported separately, with unbranded as the headline |
| What was the baseline? | Measured on the same prompts before the work began |
| How often is each prompt run? | Several times per engine, because answers vary |
If the answers are vague, the number is not ready to steer spending, however good it looks.
What I would fund
Work that is tied to a specific set of unbranded buyer questions, with a baseline taken first and a result reported on the same questions afterwards. That is slower to show progress than a dashboard that only goes up. It is also the only version that tells you whether the money did anything.
Frequently asked questions
What is the most common mistake in AI visibility reporting?
Treating a crawler visit as a citation. A bot fetching your page is not an engine naming you in an answer, but many reports show both under words like citations or mentions.
Why does the branded versus unbranded split matter for budget?
Branded prompts already contain your name, so you appear in most of them whatever you do. Spending to improve that number buys very little. Unbranded prompts are where new buyers find options, and that is where budget should be judged.
What should a good AI visibility report show?
Whether you were named or cited in answers to a fixed set of unbranded buyer questions, on named engines, measured with repeated runs, against a baseline taken before the work started.
Should we stop paying for our AI visibility tool?
Not necessarily. Most tools collect useful data. The point is to know which event each metric counts, and to base spending decisions on the ones that reflect what buyers actually see.
See how you show up today.
Five questions, answered in minutes: an estimated visibility score, the competitor most likely taking your citations, and the fix categories that matter. No card, no commitment. Prefer to talk it through first? Book a consultation with a strategist.