29

I write landing pages for a living and I refuse to ship fake countdown timers, "14 people are viewing this", or manipulative confirmshaming. But when I ask AI for "high-converting" copy it reaches straight for that playbook.

The reframe that worked: I make it lead with the single most specific, verifiable outcome the product delivers, then earn trust with concrete proof, and I explicitly ban urgency manufactured out of thin air. Conversion comes from clarity and a believable promise, not pressure.

This prompt gives me a fold I'd actually sign my name to. What's your test for whether a value prop is specific enough versus vague marketing air?

THE PROMPT
You are a conversion copywriter with a strict ethics policy. Write the above-the-fold content for a landing page. Product: {PRODUCT}. Primary user + their goal: {USER_AND_JOB}. Single desired action: {CTA_ACTION}.

BANNED (do not use, ever): fake scarcity/countdowns, invented 'X people viewing', confirmshaming ("No thanks, I hate saving money"), unverifiable superlatives ('best', 'revolutionary', '#1'), and exclamation marks.
STRUCTURE, in this order:
1. Headline: state the single most specific, outcome the user gets, phrased as a result not a feature. <= 12 words.
2. Subhead: one sentence naming WHO it's for and the concrete mechanism (how it delivers the outcome). <= 24 words.
3. Primary CTA label: a verb + the value received (not 'Submit' / 'Sign up').
4. Trust line: one piece of concrete proof (a real metric, integration, or guarantee) - if none is provided, write a bracketed {PROOF_NEEDED} note instead of inventing one.
SPECIFICITY TEST: for the headline and subhead, if a competitor could paste their name onto the same sentence and it still reads true, it's too vague - rewrite it until it isn't.
Output 3 distinct headline/subhead options at different specificity angles (outcome, pain removed, contrarian), then recommend one and say why in a sentence.
3the verb-plus-value CTA rule killed every "Submit" button on my last project. small change, real lift.story_stella 1 month ago
refusing to invent proof and emitting {PROOF_NEEDED} instead is the detail that makes this trustworthy. i've caught models fabricating a "4.9 stars from 2,000 teams" out of nowhere.onboard_owen 1 month ago
add a comment

1 Answer

27

The "could a competitor paste their name onto this?" test is genuinely the best filter I've seen for vague copy and I'm adding it to our internal rubric. As a PM I also make it output the one assumption the headline is betting on, so we can go verify it with a user interview instead of shipping a confident guess.

THE PROMPT
After recommending a headline, add: 'Riskiest assumption: <the one belief about the user this headline depends on, and how we'd validate it in one interview question>.'

Your Answer