Our SaaS homepage had the kind of headline that wins polite nods in a board deck and confuses every actual human who lands on the page: 'Empowering operations teams to leverage workflow synergies for next-gen productivity.' I know it was bad because I wrote it, with my own hands, eighteen months ago, and I'd been quietly ashamed of it for seventeen of those months. Every time I demoed the product to a real customer I'd watch their eyes skate right past the headline to the screenshot below, which is where they actually decided whether to care. The headline was doing nothing but taking up the most valuable real estate we own.
The thing that finally pushed me over the edge was a sales call where a prospect read the headline aloud — slowly, the way you read something in a language you sort of speak — paused, and asked, 'okay, so what does it actually do?' That is the single most expensive question a homepage can provoke, because it means the page failed at its one job and now a human has to do it instead. I closed the call, opened the Value Proposition Rewriter — Kill the Jargon, Keep the Punch prompt, and decided eight minutes was a fair price to undo eighteen months of cowardice.
Three rewrites, three surfaces, one pass
I gave it the old copy, our real customer pain ('ops managers lose 2+ hours a day chasing status updates across Slack, email, and spreadsheets'), and our actual differentiator ('one unified view, no integrations to set up'). GPT 5.2 returned three versions, each calibrated to a different place copy has to live and work. Version A, the hero headline: 'One screen. Every update. No chasing.' Version B, the email subject line: 'You shouldn't have to hunt for answers that should find you.' Version C, the data-led ad: 'Operations teams cut 2+ hours of daily status-chasing with a single dashboard — no integrations, no setup.' And then, crucially, a single closing line recommending I test Version C first, because it leads with a concrete, pictureable time saving that the skeptical buyer can verify against their own miserable Tuesday.
What I appreciated is that the three versions weren't just the same sentence in three outfits. They were genuinely different bets on what a visitor cares about: A bets on clarity, B bets on empathy, C bets on proof. Having all three in front of me made the decision about which to test feel like a choice rather than a guess, which is the whole point.
The banned-words list is doing half the work
I've become convinced that the most important part of this prompt isn't the role or the examples — it's the explicit list of words the model is forbidden to use. 'Seamless,' 'leverage,' 'empower,' 'synergy,' 'robust,' 'next-gen' — that was the entire vocabulary I had reached for when I wrote the original, and it's the vocabulary every tired homepage reaches for. Remove those crutches and the model is forced to say something specific, because vagueness suddenly has nowhere to hide. It's the copywriting equivalent of taking a writer's autopilot away and making them actually steer. Try it on your own draft: ban your ten favorite words and watch what's left.
- Version B became our welcome-email subject line and shipped completely unchanged
- Version A went up as the hero, with Version C running as the headline on our LinkedIn ads
- After three weeks and roughly 4,200 unique visitors, click-through to the trial page was up 14% — not a clinical trial, but enough for the whole team to agree the old copy was dead
- The model's reasoning for testing C first was genuinely sharper than the internal debate we would otherwise have had, which usually ends with whoever talks loudest winning
The uncomfortable math
Let's sit with the numbers for a second, because they're the real story. We ran a bad value proposition for over a year. It cost us trials I will never be able to count, because you don't get a notification when someone bounces off a confusing headline — they just leave and you never know they were there. And the fix took eight minutes and one prompt. The honest caveat, the part I won't dress up: the prompt did not do the hard part. We already knew the pain and the differentiator, and we knew them because we'd done real customer interviews months earlier. What the prompt did was refuse to let us be vague about things we already understood. That's a smaller job than it sounds, and a more valuable one than it sounds, all at once.
Grab the Value Proposition Rewriter — Kill the Jargon, Keep the Punch prompt on Prompt Dock and paste in your own homepage hero. If reading it back through the prompt makes you wince, good — that wince is eighteen months of quiet shame finally surfacing, and it means you're about to fix it. I pair it with my Landing Page Hero Section — Three Variations From Your Product Brief prompt whenever I want to rebuild the entire top of the page and not just swap one headline.