Why AI visibility dashboards are mostly useless
Weighing smoke: why AI visibility dashboards are mostly useless
This piece explains why AI search visibility dashboards are unreliable. Rand Fishkin's study found that asking an AI the same question 100 times yields the same recommendation list less than 1% of the time. The randomness is baked into how models generate tokens, break down queries, and select sources—it's a feature, not a bug. The only stable signal is the consideration set: top brands appear in 90%+ of responses, but their ranking is noise. The author suggests a manual afternoon check each month beats paying for a dashboard.
Why it matters: The piece cites Rand Fishkin's research comparing brand entry into recommendation candidate pools with actual rankings, and notes how stable the two monitoring metrics are.