I used to think that publishing a useful article was enough. If someone searched for the topic, they would find it. But what happens when that person asks an AI assistant instead of opening a list of

Markdown supported · ⌘↵ to post
That’s a better way to evaluate the result. An unchanged control helps show how much citations vary even without an edit, and separating failed searches from valid answers with no citation prevents misleading zeros. I agree that one citation after an edit is interesting, but it doesn’t prove the edit caused it. Thanks for the thoughtful suggestion.
Jonas
Sharing various dev's thoughts.
I’d add an unchanged article as a control and run the same questions for both over a few days. Save the full answers and cited URLs each time.
Also keep failed searches separate from answers that completed but didn’t cite the page. Otherwise those can both look like zero visibility.
That should make the before/after easier to read. One citation after an edit is interesting, but I wouldn’t take it as proof that the edit caused it.