HumanFlow is an AI product. This page says which systems do what, so you can decide whether you are comfortable with it — and so you can disclose your own use accurately where you are asked to.
The two things we run
- The humanizer rewrites text you give it. It runs on a large language model, and the text you submit is sent to our model provider over an encrypted connection solely to produce your result.
- The detector estimates how machine-like a passage reads. It runs on a model we trained ourselves, on our own infrastructure. Your text is not sent to any third-party detector.
Your text is not training data
We do not use the text you submit to train models. It is processed to return your result.
What a detection score is, and is not
A score is a statistical estimate of how predictable writing is. It is not evidence of who wrote something. Detectors disagree with one another on identical text, and they misclassify human writing — particularly by people writing in a second language. We publish the research on this, including where it undercuts our own category: detector accuracy and false positives.
What we do not claim
We do not claim to make text undetectable, and we do not publish a bypass rate. Nobody can honestly promise what a given detector will say next month. Our editorial policy commits us to that in writing.
Disclosing your own use
If you are asked whether you used AI assistance, the honest answer after using this tool is yes. Many institutions and publishers permit it with disclosure. We would rather you disclosed and kept the tool than the other way round.
Human oversight
Read what comes back before you use it. A rewrite can shift a nuance, and a score can be wrong in either direction. The judgement stays yours.
Questions: support@humanflow.pro.