Guides
How to measure an AI visibility change
A page can change and the answers can change. Those are two measurements. One does not prove it caused the other.
Keep the before sample
The report is a snapshot: the date, the engines, the questions and the coverage. A later scan is a new sample. Cached answers are not a fresh observation.
Check the page separately from the answers
You can see whether a suggested edit is on the site by reading the page. You can see whether a later API answer names you by running the questions again. ReplyScore does not currently sell a monitoring schedule that does this for you.
What not to claim
Do not turn one before-and-after pair into a percentage uplift. Models, competitors and the rest of the web also change. A fix card is a hypothesis about the site, not a forecast of mentions.