About SlopDetector

Judging content quality shouldn't require a PhD in linguistics or hours of careful reading.

How It Started

SlopDetector emerged from a simple frustration: the internet is drowning in AI-generated slop, and existing AI detectors ask the wrong question. They focus on "Was this written by AI?" when the real question should be "Is this content worth my time?"

We built SlopDetector because humans can write slop too, and AI can sometimes produce valuable content. Origin doesn't determine quality. Substance does.

What We Actually Do

SlopDetector analyzes content for quality signals — not AI fingerprints. We look for specificity, lived experience, concrete details, and genuine insight. We categorize slop into recognizable patterns: Generic Slop, Pseudo-Insight Slop, Fake Authority Slop, Wikipedia Rehash, and Wellness Slop.

Each analysis provides a judgment (not a probability percentage), a category, and human-readable explanations with highlighted evidence. You understand why something is slop, not just that it might be.

Why This Matters

AI slop is everywhere — flooding social media, search results, YouTube, Amazon, and content farms. It wastes time, degrades information quality, and erodes trust. When you can't tell if content was crafted with care or mass-produced for engagement, everyone loses.

SlopDetector gives you a vocabulary and framework to name what you're seeing. It's a cultural tool as much as a technical one — because "slop" isn't just about AI, it's about quality.

Our Philosophy

Judgment over Probability

We give you a verdict, not a confusing percentage.

Explainability over Black Box

Every judgment comes with reasons you can understand.

Human Language over ML Jargon

No perplexity scores or burstiness metrics. Just plain talk.

Cultural Relevance over Academic Correctness

We care about real-world usefulness, not research papers.

Who Builds This

SlopDetector is an independent project, built and maintained by a single developer rather than a company or a funded team. That shapes the product in ways worth stating plainly: there is no sales team, no enterprise tier, and no incentive to inflate scores to make the tool look clever. The site is supported by advertising and the detector is free, with no account and no word limit.

It also sets the limits. The dictionary reflects one person's reading of how AI writing drifts, informed by published research and by score disputes sent in over email. When someone argues convincingly that a word was flagged unfairly, that word gets re-tiered or removed. That feedback loop is the main way the engine improves.

How We Score — and What We Refuse to Do

Scoring is rule-based and runs entirely in your browser: a tiered dictionary of AI-overused words, a cliché phrase set, plus structural measures like sentence uniformity, transition abuse, type-token ratio, and whether the text commits to any concrete names, numbers, or dates. No model, no API call, no text leaving your machine. Every score can be traced back to the exact passages that caused it — which is the point of building it this way instead of asking a language model for a verdict.

What we deliberately do not do: claim to identify whether a human or a machine wrote something. Those claims are unreliable, and the published accuracy data on AI detectors shows why — false accusations fall hardest on non-native English writers. A quality score can be argued with and checked against the highlighted evidence. An authorship verdict cannot. We would rather be useful and honest about our limits than confident and wrong.

Contact

Questions or feedback? Reach us at hello@slopdetector.org