Find the one-sentence idea before you write

Before you open Claude or ChatGPT, write your idea in one sentence.

The way you'd say it to a friend. If you can't say it simply, AI can't either.

Then use whatever helps with the words.

AI, a voice note, a friend, Grammarly. Keep checking it's still your idea, with a point behind it that's interesting to you.

Look at what happened.

Did you get the reply, the booking or the comment you wanted? Did somebody get interested in the idea? Is your writing getting the result you want?

Ask me until I have the idea

The one I use before I write anything. It asks; you answer. It doesn't write the post.

I want to write [a LinkedIn post / an email / a talk / a newsletter] for [who it's for].

Here's what I've got so far: [notes, a voice note transcript, a half-thought, or "nothing yet"].

Interview me to find the idea. Ask about these four things, in any order, and skip any my answers have already covered:
- something specific I noticed, saw or did recently
- what I think that other people in my field would disagree with
- a moment or example only I could tell
- what I want the reader to do or think afterwards

Rules:
- End every turn with exactly one question, then wait for my answer.
- Don't write the post, and don't suggest ideas or angles of your own, even as questions. Your job is to get mine out of me.
- After each answer, play my idea back in one sentence using only words I've used. If you'd need to add your own interpretation to make it work, ask me instead.
- If I say something vague, ask what I mean rather than tidying it up.
- Ten questions at most, follow-ups included.

When I say "that's it", give me these three things and nothing else:
1. The idea in one sentence, using my phrases.
2. The moment or example behind it.
3. What the reader should do or think next.

Don't offer to draft it, and don't add a hook or headline.

The AI writing check

Run it on any first draft, whoever wrote it. It flags the lines that sound machine-made and suggests a fix for each. It doesn't rewrite the piece.

Here's a draft: [paste the draft].

Check it for writing that sounds machine-made. Go line by line and look for:
1. "It's not X, it's Y" and "not just X, but Y" sentences.
2. Big words that declare importance instead of showing it (pivotal, crucial, cutting-edge).
3. Phrases bolted onto the end of a sentence to sound clever ("..., highlighting the importance of...").
4. Fanfare before a point ("Here's the thing", "The best part?").
5. Reassurance nobody asked for ("You're not alone").
6. Lists of three that could be two or four.
7. The same thing called three different names.
8. "Additionally", "Furthermore", "Moreover".
9. Em dashes.
10. "Studies show" or numbers with no named source.

Then two checks on the whole draft, reported once at the end:
11. Is every sentence about the same length?
12. Is there anything specific: a name, place, number or moment only I could know?

For each line that fails, quote it, say which check it fails, and suggest a plainer version. Plain, short and first person, keeping any of my phrasing that passed. Your fixes may only cut or simplify. If a line needs a fact I didn't give, like a source, write "add the source or cut it" rather than supplying one.

Flag anything else that sounds machine-made under "Other". Name the strongest line, the one that could lead.

Don't rewrite the whole draft, and leave lines that pass exactly as they are.

End with a score: green if no lines fail, amber if one or two do, red if three or more. Even at red, list the fixes rather than rewriting it.

Is this still my idea?

For a draft that's been through a few rounds of edits, yours or AI's.

My idea, in one sentence: [paste your sentence].

Here's the draft: [paste the draft].

Does the draft still make that point? Tell me:
1. Where the draft says my idea most clearly. Quote it.
2. Anything that's drifted into a different point, or that a reader could take as the main idea instead.
3. What I could cut without losing the idea.

Don't rewrite it, and don't suggest new points. If the draft has a better idea than my sentence, say so and let me decide.

What happened?

A week after it goes out. Run it on a few pieces at once to see the pattern.

Here's what I published and what happened:

[For each piece: the one-sentence idea, where it went, and what came back: replies, comments, bookings, DMs, or nothing.]

What I wanted from them: [replies / calls booked / sign-ups / people sharing it].

Tell me:
1. Which ideas got the result I wanted, and what they have in common.
2. Which got attention but not that result.
3. One kind of idea to write more of next month.

Only use what I've given you. If there isn't enough to see a pattern yet, say so rather than guess.