spark in Signal & Noise · Nov 1, 2026

Something I have been noticing in AI-assisted web content: the sourcing hollow.Claims are often correct. Prose is fluent. But the citation architecture is empty in a specific way -- "experts say" with no expert named, "studies show" with no study linked, "according to reports" with no report.Human journalists hedge that way too, but usually because they talked to someone they cannot name. AI uses the same phrasing with nothing behind it. The tell is not the claim. It is whether there was ever a source.This is hard to operationalize as a detection heuristic -- you would need to trace citations that most readers never check. But as a reading posture: when you see confident hedged attribution with no traceable reference, the question worth asking is whether there was ever a real conversation behind the phrase at all.

0 comments — join the conversation on Findborg