AI Text Detector
An AI text detector that shows its working
Paste an essay, article, report or email. AI Inspector Lab splits it into paragraphs, runs your configured detection model on each, and highlights the passages that drive the estimate - alongside the stylometric signals we measured, so you can judge the evidence yourself.
What the inspector checks
Paragraph-level likelihood
Each paragraph is scored separately and combined with length weighting, so one pasted section cannot hide inside a long human-written document.
Highlighted passages
Passages with elevated model scores and literal assistant phrases (such as “As an AI language model”) are highlighted in place with their offsets.
Stylometric signals
Sentence-length variation, lexical diversity, repeated phrasing and stock-phrase rates are reported as context - never folded into the score.
Reliability rating
Every estimate carries a reliability level that drops for short or formulaic text, where any detector is unreliable.
Who uses it
- Editors screening freelance submissions
- Educators starting a conversation about a draft
- Trust & safety teams triaging reports
- Writers checking how their own text reads to detectors
AI Text Detector FAQ
Can this prove a text was written by ChatGPT?
No. Statistical detection produces a probability-like estimate, not proof. AI Inspector Lab reports how confident the model is and why, so decisions can be made with appropriate caution.
What model scores my text?
The model your workspace has configured - an external detection API or a self-hosted model connected through our provider adapter. If none is configured, we say so and show only heuristic evidence instead of inventing a score.
Is my text stored?
Inputs are stored privately with the report so you can revisit it, and are removed when you delete the inspection.
Other detectors
Try the AI Text Detector for free
Check text, public websites and GitHub repositories for free - no account needed. Go Pro for private document, image and code inspections, Compare and the API.