How can a B2B team tell if a GEO change actually worked?
Compare two groups of buyer questions. One group is what the change targets; the other is similar questions it leaves alone. Ask both on the same days and the same AI engines, before and after the change, and look at the difference between the two movements.
A plain before-and-after is not enough, because AI answers shift on their own. If the untouched questions moved just as much, the change cannot take the credit. The result then says one of four things: it worked, a meaningful effect was ruled out, the evidence is inconclusive, or the measurement is void.
What to fix before you start
- The questions in each group, chosen before the change ships.
- What size of change would matter to your business.
- The reading window, and no extending it after you have seen results.
How long it takes
A published page has to be crawled before AI can draw on it. WaveWise waits 14 days after the change is declared live, then reads for 7 days. If an engine changes its model between the two readings, that engine is dropped; if every engine changes, the measurement is void.
Publishing and being seen are separate questions: why shipped is not seen.
Scope and limitations
- This describes a method. It reports no result, and it does not promise that a change will move any answer.
- A small set of questions can only detect a large effect, so a small effect usually reads as inconclusive.
- Related questions can be lifted by the same page, which makes this comparison err on the cautious side.