The short answer
We ask the same question seven times a round and hand you the spread as a number. AI answers do move between runs, and that movement is one of the things the audit measures. If you asked a model twice and got two different answers, you read that correctly — one run is one run, and a single reading is a value, not a trend. Every paid audit reads citation and mention counts inside the collected runs, alongside the answer variation itself. Applicable citation rates carry Wilson score 95% confidence intervals, reported with the effective sample size — and precision labels are granted by the actual width of the pooled interval, never by wishful thinking. The final report is released after the repeated runs and the delivery-package review complete.
Why this matters
Compare single runs and a flat month reads as progress while a real gain reads as decline. Budget moves on that reading and does not move back.
The spread is reported as a number beside the result. A change either clears it or it does not.
Questions buyers read next
Where this answer comes from — Published pricesHow we measureTerms, defined