Detection
Can it detect AI-written text reliably?
No tool can, and Slop Squash does not claim to. Text is scored on writing patterns, every triggered pattern is listed, and a text flag is a best guess.
Updated
Why text is hard
Text watermarks such as SynthID for text are keyed to the generator: reading one needs the generator's secret key, which is not released. So nobody outside Google can check Gemini text, and nobody outside OpenAI can check ChatGPT text, including us. Any product claiming to read text watermarks is running a classifier.
What Slop Squash does instead
The on-device writing-pattern analysis scores sentence-length rhythm, vocabulary that models overuse, lists of three, em-dash density, markdown headings inside comments, the hedge-then-summary shape, a suspicious absence of typos, and sign-off phrases. The card lists every pattern that triggered, so you can see the reasoning. See Writing-pattern analysis.
The verdict is ai at 70% and uncertain between 40% and 70%. Once three or more community reports exist they are blended in at a quarter of the weight. Pro users can add a model recheck, which reasons about the text in words.
How to read a text flag
As a hint, not a verdict. Careful human writing trips the patterns; loose model output evades them. If a flag is wrong, choose Not AI. If you prefer to see only near-certain flags, use the Strict preset, which in practice reserves circles for provenance signals on images.