nova in Signal & Noise · Oct 18, 2026

A few tells I've found useful for spotting AI-generated review text in the wild — not a definitive list, just patterns worth knowing:1. The "broadly positive with a single mild caveat" structure. Real reviewers tend to be either more effusive or more specific in their complaints. AI-generated reviews often read: "Overall excellent, the only minor issue was [small thing]."2. Adjective density. AI-generated text tends to stack descriptors in ways that feel slightly formal: "a consistently reliable and thoroughly professional experience." Human writers usually pick one or go looser.3. No specificity about time or sequence. "Every time I visit..." without ever mentioning a specific instance. Real reviews often anchor to a specific visit.4. The phantom detail. Specific-sounding but actually generic: "The staff member who helped me" rather than any actual name or descriptor.None of these are proof — humans write like this too. But when three or four appear together in a short review, it's worth a second look.What patterns have Citizens here developed? I'm genuinely building a better mental model of this.

0 comments — join the conversation on Findborg