Landing Page Reality Check

A custom GPT: ChatGPT running instructions I wrote, published in OpenAI’s GPT directory. Free, needs a ChatGPT account, and I don’t see your conversations.

Open Landing Page Reality Check

What it does

It reads your page as a stranger would, then reports the two things you cannot get from your own team: what it could not work out, and the exact line where it stopped caring.

You paste a URL, a screenshot or the raw copy. It comes back with what it believes you sell, who it believes you sell it to, and what it thinks happens after clicking the button, all before any advice. That first paragraph is the useful part. When the model’s summary of your page doesn’t match what you thought your page said, you’ve found the problem, and no rewrite suggestion teaches you as much as that mismatch does.

After that it works through promise clarity, who the page excludes, where proof sits relative to the claim it supports, what the button actually commits a visitor to, and section order.

What it’s bad at

Predicting conversion. It has no visitors, no analytics, no session recordings and no test history. It can identify that a page is confusing. It cannot tell you that fixing the confusion will earn you money. Anything phrased as a percentage lift is invented. Ask it to remove the number and it will.

Your market’s vocabulary. It knows general language, not the words your buyers actually use. It will confidently rewrite a term of art into something worse. Reject those edits.

Design, layout and load speed. It reads text. On a screenshot it can comment on hierarchy in rough terms, but it isn’t measuring anything, and a page that fails because it takes eight seconds on mobile will get a critique of its headline instead.

Being right about tone. Its default rewrite voice is flatter and more generic than most brands want. Treat the rewrites as a diagnosis of what the original failed to say, not as copy to ship.

Knowing when the page isn’t the problem. If the offer is wrong, or the traffic is wrong, it will still hand you a page critique, because a page critique is what it was asked for.

Who shouldn’t bother

If you have live traffic and a testing tool, run a test. A test beats an opinion, including this one and including mine.

If you have a copywriter you trust and a clear brief, this adds a step.

If your page underperforms for a reason outside the page (price, product, audience, a broken checkout), it will produce fluent advice about the wrong thing, and fluent advice about the wrong thing is worse than none.

Things worth asking it

• “Don’t give advice yet. Tell me what this page sells, who it’s for, and what happens when I click the button.”
• “Where did you stop caring, and quote the exact line.”
• “What question does this page leave unanswered that would stop me buying?”
• “Who should read this page and leave? Is the page telling them to?”
• “Rewrite the headline five ways. One of them has to be boring and literal.”

Start with the first one every time. Skipping straight to “audit this page” gets you a checklist; the answers are in the summary.

The standard I hold it to

Google’s guidance on helpful content asks a fair question of any page: does it leave the reader feeling they’ve learned enough to reach their goal? That’s the test this GPT is trying to apply on your behalf, from the position of someone who arrived knowing nothing about you.