When evaluating research summaries produced by AI, percentages derived from different participant groups cannot be treated as equivalent or interchangeable. A finding drawn from a specialized cohort (such as enthusiasts) reflects only that sample's specific recruitment context and cannot serve as a reliable estimate for a broader, general customer population. Missing sources or skewed recruitment criteria must be identified before interpreting sample proportions as evidence for broad organizational decisions.