What can change between two measurements?
Six kinds, listed here in the order in which they weigh. The table comes from the code that classifies your measurements, so what is here is what you find back in your report — in the same words.
| kind of finding | what happened | where the fix sits |
|---|---|---|
| feit | An engine says something about you that is demonstrably wrong — outdated, mixed up or invented. | At the source claiming it, not with you. Which is why the address comes with it. |
| kanttekening | You are named with a caveat: good but expensive, strong but far, solid but small. | In the evidence that disproves the caveat. This is the sharpest signal there is. |
| kadering | You are framed unfavourably without a single concrete objection being made. | In the sources the engine used for it — they are listed, objection or not. |
| afwezig | You do not appear in the answer. The engine did fetch sources, and you were not among them. | Not on your own page but on the sources it did read. Without that list this is an observation without an address. |
| niet-geciteerd | Your name appears, but your own page is not the source. Someone else is telling your story. | In the form of your own page: written to be quoted, structured to be machine-read. |
| niet-gekozen | You are cited and still not recommended. The costliest of the six, because you are nearly there. | Usually in the caveat above. If there is none, this is the signal to go looking for it. |
When is a difference noise and when is it a shift?
A difference is only a shift if it concerns the same question, the same market and the same engine and holds across more than one re-measurement. If the order changes on one engine while the other three stand still, that is almost always the model choosing differently, not the market thinking differently.
What is that based on, beyond your own word?
On three measurements that are not ours. PetraLabs ran six buyer questions 600 times per engine in a single day and found a visibility figure can swing up to 30 percentage points at 21 repeats from chance alone — and that a real 10-point shift is spotted, at 50 repeats, only 17% of the time. That is why we only call a difference a shift once it sits below the baseline twice in a row.
Are you in the stable core or in the carousel?
That is a different question from “am I mentioned”, and it is the one that counts. For a large share of questions there is a stable core of a handful of names returning in every round, while everything around it rotates weekly. Two companies on exactly the same 50% score can therefore be opposite cases: one was in half the rounds and may be nowhere next week, the other has been there every time since round six.
Why do you not get an alert on every movement?
Because an alert that fires too often gets switched off within a week, and is then worth less than no alert at all. You are notified when a finding changes kind or when a factual error appears — those are the two cases where there is genuinely something to do.
What can monitoring not see?
How many people actually saw your name in an AI answer. Nobody measures that: there is no impression counter inside a chat, and any provider publishing such a figure has inferred it. We would rather call it what it is than turn it into an estimate that looks reliable.
Frequently asked questions
How often is it measured again?
That depends on your route and is on the pricing page. The rule underneath is the same for everyone: a re-measurement uses the same prompt set and the same market as the baseline, otherwise it is not a re-measurement but a new one.
Can I see exactly what changed in the answer?
Yes. Both answers are kept, so you lay the before and after text side by side, with the sources that sat under each. That is exactly the part a tool that only keeps a score can no longer show.
Do you also track my position in Google?
Not today. SEO does belong here and is already partly in the code — technique and readability are measured, Googlebot and Bingbot are verified, and search terms with their intent are used. Position tracking is on the build list, and until it exists no page claims that we track positions.
What if an engine changes model?
Then that is in the report, because every measured answer carries the model that actually answered. A shift coinciding with a model change is a shift of the engine and not of you — and that distinction is why we keep the model.
Read on
Free, no account and no card.