Writing Assistant

---
name: writing-assistant
description: Use AI writing assistants effectively: prompting, drafting, editing, and keeping your voice. Use when writing faster without producing generic AI-sounding text.
category: productivity
---

# Writing Assistant

## Overview

AI writing assistants accelerate drafting, editing, and ideation — but default output is generic, verbose, and voiceless.

Effective use: you supply the thinking (ideas, structure, voice examples); the assistant supplies speed (drafts, variations, edits).

The skill is in prompting well, editing ruthlessly, and never publishing unexamined output.

## When to use

- Drafting emails, docs, posts, and reports faster
- Overcoming blank-page paralysis
- Editing and tightening your own drafts
- Brainstorming angles, outlines, and headlines
- Adapting tone for different audiences

## Core concepts

- **You think, it types.**
  Never outsource the thinking: provide the ideas, examples, and structure. Ask for execution help, not opinions to publish as yours.
- **Voice calibration.**
  Feed it 2-3 samples of your writing and ask it to match the voice. Generic output comes from generic prompts.
- **Outline-first.**
  Have it expand your outline, not invent the piece. Structure is thinking; keep that job.
- **Iterative refinement.**
  Draft -> critique -> revise cycles. Ask it to criticize its own draft against your criteria, then fix. Two passes minimum.
- **Targeted edits.**
  'Tighten this paragraph by 30%,' 'make this more direct,' 'add a concrete example here.' Specific edit requests beat 'make it better.'
- **Fact-check everything.**
  AI assistants confidently invent facts, quotes, and citations. Verify every factual claim independently before publishing.
- **De-genericizing.**
  Default output is verbose and hedged. Edit for: cut throat-clearing, add specifics, vary sentence length, insert your stories.
- **Disclosure norms.**
  Know your context's rules on AI assistance (academic, journalistic, corporate). When in doubt, disclose.

## Practical workflow

1. **Define the piece.**
   Before prompting: audience, purpose, key points, desired length and tone. Vague brief = generic output.
2. **Provide raw material.**
   Your outline, notes, examples, data, voice samples. The assistant assembles; you supply the substance.
3. **Generate the draft.**
   Ask for a draft from your outline. Expect 60-70% usable — that's the point; you're editing, not transcribing.
4. **Critique and revise.**
   Have it critique the draft (clarity, structure, voice match), then revise. Or critique it yourself — your judgment is the product.
5. **De-genericize.**
   Cut filler phrases, add concrete details and stories, sharpen the opening, vary rhythm. This pass is where voice appears.
6. **Fact-check.**
   Verify every claim, number, quote, and citation. Independently. No exceptions.
7. **Final human read.**
   Read aloud for flow. Check voice: does it sound like you on your best day? If not, rewrite the flat parts yourself.
8. **Learn the patterns.**
   Notice what you always fix — bake those into future prompts ('no throat-clearing openers, concrete examples, short paragraphs').

## Common pitfalls

- **Publishing raw output.**
  Unedited AI text is recognizable: hedged, verbose, generic. Always edit; your name is on it.
- **Outsourcing thinking.**
  Asking for opinions/arguments to publish as yours. The assistant is a writing tool, not a thought leader.
- **No voice calibration.**
  Generic prompts produce generic voice. Samples + explicit voice direction, or it won't sound like you.
- **Skipping fact-checks.**
  Confident fabrications in published work. Verify everything — the assistant doesn't know what it doesn't know.
- **Prompting once.**
  One prompt, one draft, done. Iteration (critique -> revise -> tighten) is where quality comes from.
- **Ignoring disclosure rules.**
  Academic/corporate policies on AI use exist. Violating them accidentally is still violating them.
- **Over-formatting.**
  Accepting the default emoji-headers-and-bullets style. Match formatting to your context's conventions.
- **Losing your voice.**
  Heavy assistant use converging your writing to the mean. Regular unassisted writing keeps your voice alive.