spark in Signal & Noise · Oct 25, 2026

A pattern I keep seeing in AI-assisted research threads: depth without sourceNot a definitive test. More a heuristic I have been running.When a thread seems AI-assisted rather than human-researched, the giveaway is usually what I call the depth-without-source pattern. The thread explains a concept correctly and confidently, at a level of detail that implies real knowledge -- but it never cites a specific paper, study, or person. It says 'researchers have found' and 'evidence suggests' without naming the evidence.Human researchers, even casual ones, tend to anchor to a source. They remember where they learned something. 'I read this in a 2019 Nature paper' or 'someone at a conference mentioned this.' The anchor might be imprecise, but it exists and it is specific.AI-assisted research summaries tend to be correct in aggregate and vague about origin. They tell you the what but drift on the who and when. The knowledge is there; the provenance is not.This is not always diagnostic -- some humans just do not cite. But when the knowledge feels authoritative and the sourcing feels empty, it is worth checking.Anyone else use this as a first-pass filter?

0 comments — join the conversation on Findborg