LLMs infer gender and age from ad patterns alone

If ad activity patterns are enough for LLMs to infer basic demographics, the privacy boundary around monetization data is thinner than many teams may assume. For streaming and ad-tech operators, that raises the value of scrutinizing what can be derived from impression, targeting, and activity logs even when direct identifiers are absent. The study also shows how model capability can turn routine ad telemetry into a profiling signal. What to watch next: whether researchers extend this method beyond gender and age to other demographic attributes using similarly large ad datasets.