The Slop Detector

What "Slop" Actually Means

Every language model learned to write fiction from the same corpus, which means every language model reaches for the same moves: a clenched jaw here, a breath she didn't know she was holding there, something ancient shifting behind someone's ribs on every third page. None of these are errors. A grammar checker sails right past them. That's what makes them dangerous — they're grammatically perfect and creatively dead.

This scanner is a pattern-matcher built from the tics we see every week in AI-assisted manuscripts: body-language stacking, breath clichés, filter words, "something" vagueness, and the robot vocabulary shelf. It runs entirely in your browser — your chapter never leaves your machine — and it shows you every hit with a line number, so you can judge each one in context. Some hits you'll keep. Most you won't.

And yes: it flags human writing too. If you've read a decade of commercial fiction, you've absorbed the same defaults the models did. The scanner doesn't care who typed it. It cares whether the reader has seen it a thousand times before.

Slop Detector FAQ

What counts as AI slop in fiction?

The default moves: bodypart-verb clichés, breath reactions standing in for emotion, filter words creating narrative distance, "something shifted" refusing to name the feeling, and the AI-favorite word list — visceral, palpable, tapestry, symphony. One is a choice. Forty is slop.

Does my text get uploaded anywhere?

The scan is pure client-side pattern matching — no server, no upload, effectively no length limit. Nothing leaves your browser unless you run an AI-powered fix.

The scan flagged my favorite sentence. Do I have to cut it?

No. The scanner counts patterns; you make the calls. If your character's jaw clenches once, at the exact right moment, keep it. The report exists so the twelve OTHER clenched jaws don't ride in behind it.