Skip to content

Markdown instructions for your LLM

Fix AI Writing

Attach the guide. Add your draft. Ask for an edit.

The original writing guide: 74 editing rules, examples, and checks for factual accuracy. Download it as Markdown or copy the full text into your LLM tool.

Original file with a website credit. No signup.

By Roman Travnikov

Before and after

Cut unnecessary wording. Keep the facts and any qualifications that matter.

An example edit
Before

It is important to note that the release was delayed due to the fact that two tests failed.

After

Two failed tests delayed the release.

Both versions give the same cause of the delay.

Preserve uncertainty
Before

The onboarding change has the potential to reduce drop-off, although the sample is small.

After

The onboarding change may reduce drop-off, but the sample is small.

The possible benefit and the small sample both stay.

Use it with a draft

  1. Download the Markdown file and attach it to your LLM conversation, or copy the guide into the conversation.
  2. Add your draft and give a specific editing request. For example:
Rewrite my draft using the attached writing guide. Preserve the meaning and factual qualifications. Return the revised text.

Review the result against your source. Check facts, citations, and any change in meaning before publishing.

What the guide covers

  • Cut stock phrasing

    Find empty introductions, inflated claims, and repeated sentence patterns. Replace them with the actual point.

  • Keep the author's meaning

    Retain the intended voice. Preserve uncertainty when the evidence calls for it.

  • Check factual claims

    Check that a citation supports its sentence. Keep genuine evidence separate from invented experience.

  • Review the finished draft

    Check the full text after rewriting. The guide also explains the limits of AI-detector scores.

The Markdown file contains the full original guide, including examples of patterns to remove. Detector results depend on the checker and its input; this guide makes no promise of a particular score.