Stop Writing Like AI
Anyone publishing AI-assisted text (LinkedIn, newsletters, client docs)
Prerequisites: Module 03 helpful, not required
Last verified 2026-07-28
Learning objectives
- Recognize the main AI writing tells, especially negative parallelism ("it's not X, it's Y")
- Name the other recurring patterns: rule of three, puffery, false ranges, elegant variation, copulative avoidance, metronome rhythm
- Build a personal anti-AI style file Claude audits against
- Run the audit loop: generate, audit, edit
- Keep the banned list current as patterns drift
What's in this module
5 lessons · 6 exercises · ~5 min reading
Starts withWhy this matters now~1 minExecutive summary
You will learn the patterns that make text read as machine-written, from negative parallelism to banned vocabulary. End state is a personal anti-AI style file installed in your workflow, with one audit loop completed on real writing. After this module you can spot AI tells instantly, have Claude audit its own output against your rules, and keep your banned list current as patterns drift.
Sources
- Anthropic, "Prompting Claude Opus 5", the source for generating from positive examples rather than prohibitions, and for the fact that current models run long by default: https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-opus-5
- Claude Help Center, "Understanding Claude's personalization features", where to install standing style preferences so you are not restating them every session: https://support.claude.com/en/articles/10185728-understanding-claude-s-personalization-features
- Ruben Hassid, "It's not [X], it's [Y]": https://ruben.substack.com/p/its-not-x-its-y
Prefer the live room?
This module also runs inside our in-person bootcamps and workshops in Copenhagen.
More in Track B
Set Up Claude Cowork
How Cowork works and how to set up the folder and context files that make it effective.
Claude to Sound Like You
How to capture your writing voice, taste, and hard rules in one compact file any AI can use.
Stop Prompting Claude
Why maintained context files beat repeated prompting, and how to manage them with Obsidian.