The literary
AI‑detector
that doesn't punish Hemingway.
Generic detectors flag minimalist prose as AI. Slopsleuth runs five audits calibrated specifically for fiction — every audit names the patterns it looks for and quotes every passage it flags. No black box. See our calibration methodology for the full sample list, known limitations, and roadmap.
“It was a good fight. Not bad. Just enough. He had not expected the boy to fight at all.”
Initial calibration
Four public-domain novels.
All scored 0 / 100.
We ran four public-domain minimalist novels from 1895–1926 — Crane, Fitzgerald, Hemingway, Anderson — alongside a 1,636-word Claude-generated thriller for benchmarking. The gap is reproducible on these samples. This is a starting point, not a complete calibration set. We're actively expanding to contemporary, translated, and genre fiction — read the methodology for the full picture, including what Slopsleuth doesn't yet cover.
Tested against revision slopsleuth-app-00027-qn2,
public-domain text via Project Gutenberg. Methodology in our
Hemingway analysis.
Three audiences
Built so real authors don't get killed by an algorithm.
For authors
Audit your own draft before submission. Find the AI-prose tics you didn't know you'd absorbed — even if you wrote every word yourself. Your minimalist style won't be punished.
For literary agents
Triage the slush pile with confidence. Defensible, transparent audits — every flag explained, every passage shown. No black-box percentage you can't defend in a rejection letter.
For publishers
U.S. copyright law denies protection to AI-generated text. Catch hybrid manuscripts before contract — our voice-variance audit detects bimodal patterns whole-manuscript averages mask.
How it works
Five audits. Every flag is explainable.
No black-box statistical magic. Each audit measures a specific stylistic pattern documented in current AI prose, with thresholds calibrated against published human fiction.
Not-X Fragments
Hedge fragments like “Not because he was brave. Because his legs wouldn’t move.” The most common AI tell in modern fiction prose.
Body-Part Clusters
When the same body part (chest, jaw, throat) is hammered repeatedly within 10 paragraphs. AI reaches for physical anchors humans don’t.
Dialogue Texture
Real human dialogue is rough — fillers, restarts, mishearings. AI dialogue is unnaturally clean. We measure the absence.
Stock AI-Phrasings
“The way grief lives in objects.” “The kind of dead that meant…” Two constructions show 45–122× separation between AI and human prose.
Voice Variance
Per-chapter audit that detects hybrid manuscripts — half-human, half-AI. Bimodal patterns reveal the seams that whole-manuscript averages mask.
Why we exist
Generic detectors weren't built for fiction.
Hemingway scored 73% AI
- ✗ Trained on average internet text — minimalist prose looks "AI-like"
- ✗ Single percentage score with no explanation
- ✗ Cannot defend rejections — "73% AI" isn't actionable
- ✗ Misses hybrid manuscripts — averages hide the seams
Hemingway scored 0/100 — within human range
- ✓ Hand-tuned against four pre-1930 novels + AI samples (full methodology)
- ✓ Five transparent audits — every flag shown with the passage
- ✓ Recommended-action panel — "Read normally" / "Read with scrutiny"
- ✓ Voice-variance audit detects hybrid AI/human chapters
Pricing
Plans for writers, agents, and agencies.
Start at $7/month with a 7-day free trial. Cancel anytime — annual subscribers get full term, no early-cancel fees.
Don't let the next great voice
die at the gate.
Run a free sample audit in 30 seconds. No signup. No credit card. Just the engine.
Try Slopsleuth free