Write Landing Page Copy That Says Something
Produce a full landing page — headline, subhead, three benefits, objection handling, call to action — for a real product. The method is research-first: you interview the customer before you write a word.
Why this one
Generated landing pages fail because they describe the product. Pages that work describe the reader's problem accurately enough that they feel understood. That requires input the model does not have, so most of the work is supplying it.
Steps
Write the customer down
One paragraph: who they are, what they are currently doing instead, what specifically annoys them about it, and what they would have to believe to switch. Guess if you have to, but write it down — vague audience input produces vague copy, reliably.
Generate twenty headlines and reject nineteen
Ask for twenty across different angles — outcome, problem, contrarian, specific-number. Volume is the point: the first three will be generic because generic is the average of the training data, and the interesting ones sit at position eleven onwards.
PromptProduct: [what it does, one sentence] Customer: [your paragraph] Current alternative: [what they do today] Write 20 landing page headlines. Vary the angle: outcome-led, problem-led, contrarian, specific-number, and direct. No question marks. No colons. Nothing longer than twelve words. Do not explain them.
Turn features into consequences
For each of your three main features, ask "and therefore?" three times. "Real-time sync" becomes "your team sees the change immediately" becomes "nobody works from a stale copy" becomes "you stop finding out about mistakes a week later". The third answer is the benefit worth putting on the page.
Handle the objection you are avoiding
Ask the model to list the five reasons someone would read your page and not sign up, then answer the strongest one directly in the copy. Pages that pretend the objection does not exist read as evasive, and the reader has already thought of it.
Read it as the customer
Give the model your customer paragraph and the finished page and ask it to react in character, saying which line made it sceptical. This will not replace real user testing, but it catches the obvious misses before anyone else sees it.
You should end up with
A complete landing page draft with a documented rationale for the headline.
Done when
- The headline names a consequence, not a feature
- The strongest objection is addressed on the page
- Every benefit survives the "and therefore?" test three times
If you want to go further
- Write a second version for a different customer segment and compare
- Build the page from a landing page template and put the copy in it