Before assuming a bug — though sometimes it is one, and we would rather hear about it — check whether the two numbers are measuring the same thing. They usually are not.
The usual causes
Different date ranges. A card showing "last 28 days" and a chart showing "last 7" will disagree, correctly.
Different denominators. "Prompts naming you" counts prompts. "Citations" counts links inside answers. One prompt can produce eight citations, so the numbers are not comparable and were never meant to be.
Unique things versus total things. "22 cited sites" counts distinct domains. "30 citations" counts individual links. Both true, different units.
A filter you forgot. Charts that follow a filtered table show the filtered set. If a filter is active, the page says so — but it is easy to set one, navigate away, and come back.
Latest run versus all runs. A per-prompt status reflects its most recent run. A trend sums every run in a day. On a day with several runs those numbers are different by design.
Search Console lag. Google is a day or two behind on recent dates, so a comparison spanning yesterday can look like a drop that has not happened.
When it is worth reporting
Report it when two numbers that genuinely measure the same thing, over the same range, disagree. That is a real bug and we want it.
Include:
- The two screens and what each said
- The date range and any filters that were active
- The project
Open the ticket from inside the app so the project is attached automatically. That is usually the difference between us reproducing it in five minutes and asking you three follow-up questions.
The honest note
We have shipped this bug before. Two cards on the AI Tracker once disagreed about the same word — one counted prompts where the brand was named, the other counted prompts where anything was cited, and both were labeled "visible". If a number looks wrong to you, it is worth asking.
Next
- Reading AI visibility: the five states
- Reading your site audit results