Private by design: runs entirely in your browser
Was this written by AI?Get an honest estimate.
AI Text Check analyzes writing patterns with its own detection model and estimates how likely a text is to be AI-generated. No sign-up, and your text never leaves your device.
Your result will appear here
Paste text and press Analyze. You'll get an estimated AI probability, not a final judgement.
Analyzed on your device. Your text is never uploaded.
How it works
A compact neural network, trained by us on tens of thousands of labelled human and AI texts, reads your text and estimates the probability that it was machine-generated.
Stays on your device
The model is downloaded once and runs in your browser with WebAssembly. Nothing you paste is uploaded.
Our own model
A transformer classifier fine-tuned on large public research datasets and real student essays, covering 30+ AI models and many writing styles.
Estimates, not verdicts
Long texts are scored section by section and combined. Results in the middle range are shown as uncertain instead of forcing a call.
What to expect from results
AI Text Check is built to identify AI-generated writing while avoiding overconfident results. When the signals are not clear, it says so instead of guessing, so genuine human writing is not labelled as AI without good reason.
- Longer English text gives the detector more to work with and leads to clearer results.
- Heavily edited, paraphrased or mixed human/AI writing can be harder to identify.
- Some human writing, such as very formal or formulaic text, can look AI-like.
- No AI detector is perfect. Treat the result as an indication, not as proof.
Frequently asked questions
Can this prove that someone used AI?+
No. No detector can prove authorship. AI Text Check gives a probability based on writing patterns. Human writing can look AI-like (for example very formal, simple or non-native English) and AI text that has been edited or paraphrased can look human. Treat the result as one signal to start a conversation, never as evidence on its own.
Is my text stored or sent anywhere?+
No. The detection model is downloaded to your browser and the analysis runs on your device. The text you paste is not uploaded to our servers or to any third party, and we don't keep a history.
Which AI models can it detect?+
It was trained on writing from many models, including ChatGPT, GPT-4, Llama, Mistral, MPT, OPT, Flan-T5 and others, and on human writing such as student essays, news, stories, Reddit posts and Q&A answers. It works best on English text of 120 words or more. It is deliberately cautious: when it isn't confident it says "uncertain" rather than risk calling a person's writing AI. Newer models, edited AI text and text rewritten by paraphrasing tools often land in the uncertain range.
Why does the first analysis take longer?+
The first time, your browser downloads our detection model (about 60 MB). After that it is cached, so later analyses start immediately, even offline.
What does "Uncertain / mixed" mean?+
The model's estimate is in the middle range. This often happens with AI text that a person edited, human text polished by AI, or styles the model rarely saw. In these cases no confident call is possible.