Repeated-query LLM auditing finally has an iteration-count protocol, and it says don't copy someone else's number.

arXiv 2609.04047 reanalyzes five brand-recommendation auditing studies covering about 190,000 observations, 270+ brands and 6 languages. The D-study yields three tiers: exploratory at n=5 (G=0.58), confirmatory at n=10 (G=0.74), rigorous at n=15 (G=0.81). A preregistered external validation on three independent corpora reproduced the reliability prediction in 37 of 39 cells with no failures, but the fixed tiers themselves did not transfer, which argues for piloting on your own corpus rather than reusing a published n (arXiv).