Clankprint checks whether a language model was involved in writing a text. Paste a draft or upload a file, and get one calibrated probability plus the paragraphs that drive it.
The headline is the likelihood that a language model was involved. Below it, every paragraph is shaded by how strongly it carries the signal, so you can see where to look.
over 99%
Likelihood that an LLM was involved in writing this text.
Likely written with AI (strong evidence)
Where the signal is · 5 of 5 paragraphs likely mostly AI
A draft, or a PDF, DOCX, Markdown or plain-text file. Up to 2,000 words in the free beta.
The likelihood that a language model was involved in writing the text. It is not the share of the text written by AI.
Paragraphs are shaded by signal strength and labelled with what each tier has been measured to mean.
Clankprint never states certainty the engine didn't give: the top of the scale reads “over 99%”. A result is a likelihood about a text, not a judgement of the person who wrote it, and it carries no guarantee against determined paraphrasing.
Create an API key on your account page and call the same engine from a script. Web and API return the same verdict. API access isn't part of the free beta yet.
curl https://clankercheck-api.vercel.app/v1/checks \
-H "Authorization: Bearer cck_…9f2a" \
-H "Content-Type: application/json" \
-d '{"text": "…"}'