SlopTotal
Self-hosted AI text detector that runs 23 detection engines on your own CPU and returns one calibrated score.
Scan text, files or URLs on your own machine and read each detection engine's vote.
SlopTotal is a self-hosted AI text detector that scores pasted text, PDF or Word files and web pages with 23 engines at once. It suits teachers, editors and researchers who want to inspect how a verdict was reached and keep documents off a vendor server.
What it is
It runs neural classifiers, Binoculars, Fast-DetectGPT, GLTR and perplexity tests in parallel and combines their votes into one calibrated score. You can call it through a JSON API, and the README documents a per-paragraph heat map through that API.
Who it's for
It suits anyone who can run a Docker container and wants an inspectable detector. It does not suit people who need a vendor-backed service, and the README says essays by non-native English writers are still flagged too often.
Key features
23 detection engines
DeBERTa and RoBERTa classifiers, Binoculars, Fast-DetectGPT, GLTR and statistical tests run in parallel.
Calibrated ensemble score
Engine votes combine into one verdict, and you can inspect each engine's result.
Text, URL and file input
Paste text, scan a web page, or upload .pdf, .docx, .txt or .md files.
Published accuracy data
The README reports results on its SlopBench corpus, with the harness and raw scores.
Site check for AI app builders
Looks for fingerprints that tools such as Lovable, v0 and Bolt leave in deployed sites.
JSON API and Chrome extension
Call it from scripts, or mark AI-looking results in Google Search with the companion extension.
Up and running in 3 steps
Our honest verdict
What's great
- Every engine's vote is visible, so you can see why a text scored high instead of trusting one number.
- Runs on CPU with a single docker run command; a GPU is optional.
- Reports are deleted after 30 days and no third-party AI APIs are called, per the README.
- The README publishes where it fails, including false-positive rates on human text.
Where it falls short
- Loaded models need about 8.3 GB of RAM, so a small VPS will struggle.
- The README says essays by non-native English writers are still flagged.
- Languages beyond English are experimental or unreliable, per the README.
- Scores are evidence for a human reviewer, not proof of AI authorship.
Pricing & license
SlopTotal is free under the MIT licence. The project also runs a free live demo at sloptotal.com; otherwise the only cost is running the container yourself.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Does SlopTotal send my text to another service?
The README says it uses no third-party AI APIs and runs on your own hardware, so scanned text stays on your machine.
Can it run without a GPU?
Yes. The README says CPU-only is fine and a GPU is optional, but loaded models need about 8.3 GB of RAM.
Is it accurate on student essays?
The README publishes accuracy and false-positive figures from its own SlopBench corpus and warns that non-native English essays are still flagged, so treat scores as evidence only.
