AI Humanize

The homepage keeps six purpose-written drafts, from a work email to an academic paragraph, next to the actual model output and a follow-up revision. The site says the samples are fictional scenarios shown without manual edits, and it labels when each was generated (October 6, 2026 for the work email). That lets you judge the rewriting style before pasting anything of your own.

In the editor, the original and revised text sit side by side. Manual edits are free, you can ask for another pass with a specific instruction, and signed-in rewrites are saved with their versions. TXT and Markdown files can be imported and exported, and the draft keeps its original language by default.

The FAQ answers the obvious question directly. There's no detector score or bypass guarantee, you should follow any disclosure or authorship rules that apply to your work, and you should check every rewrite yourself. The academic use-case page takes the same line. It's about keeping named terms, study populations and statistical meaning exact, and it says those checks stay with the researcher.

Credits follow token usage for both input and output, so a long draft costs more than a short one, and the site doesn't turn credits into a fixed number of rewrites. Every plan has the same model capacity, up to 922,000 input tokens and 12,800 generation tokens. Basic is $9.90 a month for 120 credits and Pro is $19.90 for 300, annual billing is half price, and a $4.90 pack of 50 credits never expires.

The privacy policy, dated October 10, 2026, says text goes to the language model through APIMART and that the operator can't promise the provider keeps nothing. Anonymous results are kept for up to 24 hours, and signed-in writing stays in your history until you delete it. The service is run by an individual operator based in China.