Prompting for Usable Drafts
The Self-Critique Pass
Asking a model to review its own draft is the cheapest quality upgrade in prompting — and the only one with a hard number on facts: 7.5% to 77.5% accurate citations.
7 cited sources
Guides on prompting, editing, and structuring copy with tools like ChatGPT and Claude — for bloggers, marketers, and founders who write for a living.
New guides on writing and editing with AI — every claim tied to a source you can open and check.
Prompting for Usable Drafts
Asking a model to review its own draft is the cheapest quality upgrade in prompting — and the only one with a hard number on facts: 7.5% to 77.5% accurate citations.
7 cited sources
Prompting for Usable Drafts
Writing English as a Non-Native
Writing English as a Non-Native
Writing English as a Non-Native
A corrected sentence is still yours; a rewritten one belongs to the model. For non-native English writers that line decides both authorship and whether you ever learn to write without the tool.
5 cited sources
Writing English as a Non-Native
AI feedback reliably fixes two things in second-language writing — grammatical accuracy and lexical fluency. It does not teach you to write, and the difference matters.
6 cited sources
Fact-Checking AI Output
Yes — a follow-up verify prompt measurably improves citation accuracy (7.5% to 77.5% in one study). No — self-checking still leaves a fifth wrong, because a model can't reliably audit itself.
7 cited sources
Four kinds of posts, one goal: copy that reads like a person wrote it.
Step-by-step breakdowns of prompts that actually produce a usable first draft — not just “write me a blog post about X.”
How to cut the filler, fix the rhythm, and turn a generic AI draft into something worth publishing.
Honest notes on where ChatGPT, Claude, and other writing tools are actually good — and where they fall apart.
Ways to fit AI into an existing writing process without it taking over the parts you're actually good at.
Not a magic prompt. A process — four steps, repeated every time.
Audience, goal, tone — write it down before you open the AI tool. Vague input is why most AI drafts sound the same.
Outline first, then a rough draft, then a pass focused only on voice. Trying to get a finished piece from one prompt rarely works.
Read it out loud. If a sentence wouldn't come out of your mouth, rewrite it — even if the AI wrote something “correct.”
Save the prompts and edits that worked. Writing with AI gets faster once you stop reinventing the process every time.
Not for being AI-assisted — Google's guidance is about quality and usefulness, not how a draft got written. Thin, unedited AI output gets buried for the same reason thin human-written content does.
Depends on the task. Claude tends to hold a consistent voice better over longer pieces; ChatGPT is fast to iterate with for outlines and short-form copy. Most people writing regularly end up using both.
Feed it your own examples instead of asking it to “sound professional.” The tool mirrors whatever tone you show it — generic input gets generic output.
For writing specifically, yes — a well-structured prompt with context, audience, and constraints saves more editing time than any single “trick” phrase.
Both. Long-form guides and ad or landing-page copy need different techniques with AI, and this site covers each separately rather than treating “writing” as one skill.
Send a message if something in a guide doesn't fit your use case, or if there's a tool or workflow you'd like covered.