AI wrote the first draft. Make it sound like you.

Limits

What markhuman will not promise

Most AI drafts fail as writing in the same few ways: even sentence length, signposted transitions, a closing paragraph that restates the opening. A thesaurus pass leaves all of that in place. markhuman regenerates the passage — rhythm, word choice, and the joints between sentences — then scores the result with the same detector that powers the rest of the site. Only a draft that clears that check is returned.

Nothing in the pipeline verifies that your meaning survived — read numbers, names, and dates against the original before you use the output. False flags cluster on formal prose and fluent second-language English; missed detections cluster on heavily edited machine drafts.

The free allowance is a daily pool, not a trial: no card, no email gate, no blur, and signing in raises the cap and nothing else. Short and mid-length passages are where the rewrite works best; long formal documents are the hard case for every automated rewriter, ours included.

The humanizer is for drafts that already say what you mean and still read like a model wrote them. The detector shows which sentences carry the signal — in your work, or someone else's. Classrooms and content teams use the same plans on the pricing section; /contact stays in the footer for anything we have not put on the page.

A ChatGPT, Claude, or Gemini draft you actually agree with: the argument is yours; the scaffolding is not. Signposted transitions, even paragraph mass, a restating close — that is what the rewrite removes. It will not invent a view you have not taken.

An essay, email, or post before it leaves your hands: run the detector on the original, humanize, then re-check. The gate is our own detector, and a pass here is evidence about our scanner only — not a translation of GPTZero or Turnitin.

A submission, pitch, or page you have to review: sentence shading shows which lines drove the call. AI-Assisted is its own verdict because most real documents sit in the middle. Do not use a report as the sole basis for a grade, a rejection, or a hiring decision.

Not: a way to hide work you did not do. Schools, employers, and publishers have their own rules; cleaner cadence does not make writing yours. If you cannot talk through the claims without the draft in front of you, write the thing.

We do not publish pass rates for GPTZero, Turnitin, Originality.ai, or anyone else. Detectors disagree often enough that a number we have not measured would be a claim, not a result. Every humanizer draft is scored by our own detector, and only a passing one is returned. If you need a third-party read, run that scanner yourself.

The detector's free tier is 50 scans a day at up to 2,500 words each; signed-up free accounts add 20,000 detector words a month. The humanizer's free pool is 2,000 words a day without an account and 10,000 signed in, up to 1,200 words per rewrite. Unused words do not roll over; pools refresh the next UTC day. Signing in raises the caps and nothing else — no blur, no partial output, no card field in the free flow. If no candidate draft clears our detector, the words are refunded. Pasted text is sent to our inference service to be rewritten or scored. It is not sold, shared, or used to train a model.