Check whether a text reads like it was written by AI tools such as ChatGPT, Gemini or Claude. Paste at least 40 words and get an AI likelihood score, with the sentences that read most like AI highlighted. Free, with no sign-up.
Strong AI signals Some AI signals Unmarked: reads human
AI detectors can be wrong. Formal, polished or heavily edited human writing can score high, and AI text that has been rewritten can score low. Never treat this result as proof on its own.
How to check text for AI
- Paste at least 40 words (up to 2,000 characters, about 350 words) into the box.
- Click Check for AI.
- Read the score and the highlighted sentences: red shows strong AI signals, amber shows some.
- For longer documents, check one section at a time.
How to read your result
| Result | What it means |
|---|---|
| Reads like human writing (under 20%) | Few AI signals were found. |
| Mixed signals (20% to 79%) | Parts read like AI and parts read human. This is common for AI text that someone edited, and for very polished human writing. |
| Strong signs of AI writing (80% and above) | Large parts of the text read like AI-generated writing. |
How accurate is it?
The checker uses a trained AI-detection model from Sapling, a writing-tools company. In our own tests it picked up every ChatGPT-written sample we tried and did not flag casual human writing such as forum comments. It was much less reliable on formal, encyclopedia-style human writing, which it often scored as AI.
That is true of all AI detectors: none of them can prove who wrote a text (see are AI detectors accurate? for what the research shows). Treat the score as one signal, together with what you know about the writer, their drafts and their usual style. Teachers in particular should never rely on a detector score alone.
Why human writing can score high
- Formal, textbook-style writing with no personal voice.
- Text that was run through grammar or rewriting tools.
- Writing that follows a template, such as standard cover letters or product descriptions.
- Very short texts, which give the detector little to go on.
Privacy
To score your text, it is sent securely to Sapling, our AI-detection provider. SpellMistake does not keep a copy beyond a short-lived cache that avoids re-checking identical text within 24 hours. See the Privacy Policy. Please do not paste confidential information.
Curious which words make writing look machine-made? Read our guide to AI words to avoid.
Related tools and guides
- AI detection guide: everything in one place
- AI humanizer
- Paragraph rewriter
- Free grammar checker
- How to paraphrase
- Essay checker
- Are AI detectors accurate?
- Is GPTZero accurate?
- Does Canvas detect AI?
- Is Grammarly's AI detector accurate?
- How accurate is Turnitin's AI detector?
- Is QuillBot's AI detector accurate?
FAQ
Is this AI detector free?
Yes. You can run up to 10 checks a day for free, with no account.
Can it detect ChatGPT?
Yes. It is trained to spot text from popular AI models, including ChatGPT. No detector is perfect, so read the result with care.
Why is my own writing marked as AI?
Formal, polished or template-like writing can look like AI text to a detector. A high score is not proof that you used AI.
How long can my text be?
Up to 2,000 characters per check, which is about 350 words. Check longer texts in sections.
Do you store my text?
Your text is sent to our detection provider to be scored. We keep a temporary copy for up to 24 hours so the same text is not checked twice, then it is deleted.