# SlopScanner

> Free browser-based tool that scores text 0-100 for AI-generated writing patterns.

SlopScanner checks prose for patterns statistically overrepresented in LLM output. It is a browser port of SlopGuard v0.3.1, an open-source Python linter by Eric W. Tramel, Ph.D.

## What it does

- Matches ~95 compiled regex patterns against input text
- Detects overused words, stock phrases, structural patterns, rhythm anomalies, and concentration of repeated tics
- Produces a score from 0 (saturated with AI patterns) to 100 (clean human writing)
- Runs entirely client-side. No text is transmitted anywhere.

## Scoring bands

- 80-100 Clean: Reads like a human wrote it.
- 60-79 Light: A few patterns detected. Worth a second look.
- 40-59 Moderate: Noticeable AI fingerprints. Revise before sharing.
- 20-39 Heavy: Clearly formulaic. Needs significant rewriting.
- 0-19 Saturated: Wall-to-wall slop. Start over.

## Detection categories

- Words and Phrases: ~80 overused words and ~50 stock phrases common in LLM output
- Structure: Bold-header blocks, bullet runs, triadic lists, blockquotes-as-thesis, closing aphorism detection
- Rhythm and Density: Sentence length variance, em dash density, elaboration colon density, extreme run-on sentences, copula chain density, paragraph length uniformity
- Concentration: Repeated use of the same pattern costs exponentially more

## Usage

Visit https://slopscanner.com, paste text or upload a file (.txt, .rtf, .docx, .pdf), and click Scan for Slop.

## Attribution

Scoring engine: SlopGuard by Eric W. Tramel, Ph.D.
Source: https://github.com/eric-tramel/slop-guard
License: MIT
Blog post: https://eric-tramel.github.io/blog/2026-02-18-slop-guard/