AI Content Forensics

Analyse suspicious text with 23 detection engines.

Paste a URL or text. Get a detailed forensic report from 23 independent AI detectors — statistical, neural, and linguistic — all running locally.

Free & private. All analysis runs locally on our servers.

Detection Engines

23

Each engine uses a different methodology. Results are weighted and combined into a single forensic score.

TM TMR Detector neural RM ReMoDetect neural BN Binoculars statistical FD Fast-DetectGPT statistical PP Perplexity statistical XP Cross-Perplexity statistical FS Fakespot neural E5 E5-Small embedding BT BERT-tiny RAID neural OA OpenAI Detector classifier CG ChatGPT Detector classifier DL Desklib DeBERTa neural SN SuperAnnotate neural GL GLTR statistical LR Log-Rank statistical
DE DivEye statistical
BU Burstiness linguistic
LM Linguistic Markers linguistic
SA Structural Analysis linguistic
VR Vocabulary Richness linguistic
FP Formulaic Patterns linguistic
RU Readability Uniformity linguistic
SH Sentiment & Hedging linguistic

How It Works

01

Submit

Paste text directly or provide a URL. We extract the article content automatically.

02

Analyse

23 engines run in parallel — neural classifiers, statistical tests, and linguistic heuristics.

03

Report

Results stream in real-time. Each engine votes independently. A calibrated, weighted score produces the final verdict.

Reference

What is “slop”?

Slop is internet shorthand for low-effort, AI-generated content published at scale — articles, comments, reviews, and social posts churned out by language models with little or no human editing.

The term emerged in 2024 as LLM-generated text began flooding search results, product reviews, LinkedIn feeds, and news sites. Unlike thoughtful AI-assisted writing, slop is characterised by its formulaic structure, hollow confidence, and lack of genuine insight.

SlopTotal doesn’t judge all AI use. Humans editing AI drafts, using AI for brainstorming, or collaborating with AI tools is not slop. What we detect is the telltale fingerprint of text that was generated wholesale and published without meaningful human involvement.

Common slop signals

  • Formulaic phrases “In today’s rapidly evolving landscape”, “It’s important to note”, “delve into”
  • Rigid structure Numbered lists, “Firstly / Secondly / Moreover” stacking, five-paragraph essay format
  • Hedging & filler Excessive qualifiers, hollow conclusions, saying a lot while conveying little
  • Statistical fingerprints Low perplexity, uniform burstiness, predictable token distributions

Reading Your Score

The overall score represents the probability that the text was AI-generated, calibrated across all engines.

0–25
Likely Human Low AI signals across engines. The text shows patterns consistent with human authorship.
25–50
Probably Human Minor signals detected. Could be formal writing, AI-assisted editing, or a false positive on polished text.
50–65
Mixed / Uncertain Engines disagree. May be AI-assisted, heavily edited AI output, or formal human writing that triggers classifiers.
65–80
Likely AI Strong signals from multiple engines. High probability of substantial AI generation with limited human editing.
80–100
Almost Certainly AI Broad consensus across engines. The text has clear statistical and linguistic markers of unedited AI output.