A new IQRush paper, due for release next week, argues that AI visibility rankings are often unreliable due to statistical noise, as generative models produce different responses each time. The paper, by IQRush co-founder Ron Sielinski, provides a method to distinguish genuine differences from fluctuation, showing that no fixed amount of data can definitively settle the question. A separate team published a similar finding in April. The paper states that for a ranking to be reliable, the order must stop changing and the top sites must be well apart, with the difference between them exceeding the margin of error. In 30 platform-topic tests, the number of answers needed for both conditions ranged from 33 to 94, counting only answers with citations. Three out of 30 tests did not reach this point even after 125 questions, all on SearchGPT, where top sites were too similar to tell apart.