Paste any text and see, sentence by sentence, what was likely machine-written. Detects ChatGPT, Claude, Gemini, Llama and humanizer tools.
See exactly which passages look machine-written, not just a single score.
Trained on paraphrased and rewritten text from popular bypass tools.
Estimates which model family likely produced the text.
Check a whole class or submission queue at once. PDF, DOCX and Google Docs.
Works inside Canvas, Google Classroom, Moodle and Blackboard.
Export a PDF with highlights to share with students or authors.
Tested on 10,000 human and AI documents across essays, news, reviews and fiction.
Technical report →From classrooms and editorial desks to legal practices and data teams, explore where an authorship review fits.
Review student writing for possible AI assistance, then use the passages as a starting point for a conversation about drafts, revision, and the learning process.
What you might review
“The highlights changed how I talk to students. We look at the passages together instead of arguing over a number.”
“We screen every freelance submission now. It takes seconds and it hasn’t flagged a single writer we trust.”
“The API was a one-afternoon integration. Fake reviews dropped noticeably in the first month.”
300,000 words per month
3,000 word credits
One personal workspace
Paste text, upload a file or import a link
1,500,000 words per month
15,000 word credits
One personal workspace
Paste text, upload a file or import a link
300,000 words per month
3,000 word credits
Two seats · shared allowance
Paste text, upload a file or import a link
Start with a free demo workspace and 100 credits. This prototype simulates plan changes and checks. No payments are collected; MELD, API access and integrations are coming later.
On our public benchmark it flags about 1 in 10,000 human documents by mistake, and catches over 99% of AI-generated text.
Yes. The model is trained on output from common paraphrasing and humanizer tools.
More than 20, including Spanish, French, German, Portuguese, Chinese and Japanese.
Free checks are deleted after 24 hours. Institutions can set their own retention policy.
No. Treat results as evidence to start a conversation, alongside drafts, version history and the writer’s own account.