How to edit an AI draft without pretending you wrote it

An AI draft is a pile of plausible sentences. It is not your research, your customer, or your liability. The useful move is not to make the pile “undetectable.” Detectors are inconsistent, and sneaking a model’s homework into a class or a journal is still dishonest when the detector fails.

The useful move is to treat the draft the way you treat a junior intern’s first pass: keep the structure if it helps, replace every claim you cannot defend, and write the sentences only you could write.

A pass I use on every AI draft

Read it once without touching it. If you cannot say what the piece is arguing, the draft is a cloud. Start over with a tighter prompt or an outline you wrote yourself.

• Delete the opening that begins with “In today’s fast-paced world” or “It is important to note.”
• Mark every fact. Keep only the ones you have a source for. Invented numbers and “studies show” with no study go.
• Replace generic examples with the one case you actually have — a PDF you cleaned, a client limit, a bug you hit.
• Vary sentence length on purpose. Machines default to a medium sentence with a comma in the middle.
• Add the disagreement. If two tools count words differently, say so. Balanced mush is a model habit.

What TextFixer’s rewriter is for

The AI Humanizer and related rewrite tools are for rhythm and word choice after you already stand behind the facts. They are not a way to submit machine text as your own academic work, and they cannot promise that a detector will go green. I will not market them that way.

If a school, a publisher, or an employer asks whether you used a model, answer them. If the answer would get you in trouble, do not use the model for that piece.

When not to use a rewriter at all

Legal filings, medical advice, and anything you will swear to should be written by the person who is accountable. A model can suggest a clearer sentence. It cannot take the stand.