Services

Refining AI-assisted drafts

Editing AI-assisted text so it’s accurate and on-brand, and reads as though a person was responsible for it.

$60 per hour, or $70 for medical, legal, and technical content

A model can produce a draft in seconds, but it doesn’t know your organization and can’t take responsibility for the text, so the draft still needs someone who understands what it’s for.

We edit AI-assisted text the same way we edit any other draft: by reading it as your audience will, and fixing what gets in the way. The difference is that AI drafts fail in specific, predictable ways, and we know where to look.

What we do to the draft

Restore the voice. Model output has a certain register: fluent and anonymous. It hedges where you’d commit, and it reaches for the same constructions over and over. We make sure your organization’s tone survives, which can mean either cutting or wholesale rewriting, depending on what’s in front of us.

Check every claim and every source. Fabricated citations are common in AI-assisted work, and the consequences can undo the reputation you spent years building. We check that each source exists and that it’s described accurately. Where something can’t be verified, we flag it in the margin and leave the decision to you.

Cut the padding. AI drafts are long because length is cheap. Many can be cut down substantially without losing a single idea.

Make the argument airtight. Generated text often states positions without supporting them, and moves between points without connecting them. We tell you when there are holes in your reasoning that only you can fill.

Fix the mechanics. Grammar, usage, punctuation, style-guide compliance, consistency of terminology, and the formatting of references and cross-references.

Where we stop

We don’t take work aimed at lowering AI detection scores, and we won’t tell you that we can influence one.

Detectors aren’t reliable: they flag careful human writing as machine-made, and they pass generated text as human (Gonçalves & Gonçalves, 2026; Weber-Wulff et al., 2023). No editor can honestly promise you a number, so we don’t offer one. If the brief is to get a document under a threshold, we’re not the right service for you, but we can promise that your document will read as clear, natural, human writing.

We also don’t run your document through AI or plagiarism checkers of any kind. Our editors judge writing the way your readers will: on the page.

How it works

We work in Microsoft Word with tracked changes, so all of our changes are visible and reversible. Where a decision is yours to make, it goes in a margin comment rather than into the text.

You get the edited document back by the agreed deadline, with tracked changes and comments. We don’t deliver clean files, with all changes and comments removed, on purpose: a central part of the process is processing our comments and changes. Without that involvement from the author, the text becomes ours, not yours.

What it costs

Billing is hourly at our standard rates. If you’re working to a fixed budget, tell us and we’ll adjust the scope to fit.

References

  • Gonçalves, B. F., & Gonçalves, V. (2026). Tools for detecting and humanizing AI in academic work: An exploratory study. In A. Rocha, C. Ferrás, & L. Loo (Eds.), Information technology & systems (Lecture Notes in Networks and Systems, Vol. 1965, pp. 588–597). Springer Nature. https://doi.org/10.1007/978-3-032-25652-2_48
  • Weber-Wulff, D., Anohina-Naumeca, A., Bjelobaba, S., Foltýnek, T., Guerrero-Dib, J., Popoola, O., Šigut, P., & Waddington, L. (2023). Testing of detection tools for AI-generated text. International Journal for Educational Integrity, 19, Article 26. https://doi.org/10.1007/s40979-023-00146-z

Send us the draft

Whatever state it’s in, and however the first version was written. We’ll agree the scope with you before quoting.

Get a quote