AI tell · grade S
Vague attribution: "studies show" and "experts argue"
Vague attribution is a claim propped up by an authority that is never named: the research that reportedly shows something, the experts who reportedly agree. It is a common way for a model to sound evidenced without carrying evidence. Readers flag it because a real citation names a real source, and the absence of one signals that the authority was invented to lend weight. The fix is to name the source or drop the appeal and make the claim on your own.
What it is
A sentence that borrows authority it cannot produce. It gestures at a body of work or a class of experts, states what they supposedly hold, and moves on before anyone can check. The gesture is doing the persuading, and there is nothing behind it.
How often it shows up
Both legacy style guides we cross-checked flag the sourceless appeal as a weasel construction, and our social-sentiment review confirmed it as a recognized tell. The checklist grades it S, socially confirmed and skill-corroborated.
Why readers flag it
A careful reader wants the source. When the sentence names none, the appeal reads as decoration rather than proof, and a page that leans on it repeatedly reads as a model padding thin claims with borrowed credibility it never had.
The fix
Name the study, the author, or the number, or make the claim yourself and own it. If you cannot find a real source, that is a sign the claim needs softening to what you can actually stand behind.
Examples
Research consistently shows that multitasking reduces productivity, and experts agree it harms focus.
A 2009 Stanford study found that heavy multitaskers performed worse on tests of attention than people who focused on one task.
Questions
Is citing research always a tell?
No. A real citation with a name, a year, or a number is exactly what good writing does. The tell is the appeal with no source behind it, which a model uses to sound evidenced without being so.
underslop marks this tell and the rest of the consensus set in any draft you paste, using the same open lint that scores it here. The free tier caps at 500 words a run, and your text is held for 0 seconds.