Free landing page roast (2026)

We fetch your page the way a browser would and extract what a visitor actually processes in their first five seconds: the title, the meta description, the heading hierarchy and the visible body copy.

We do not execute your JavaScript. If your marketing page is a client-rendered app with an empty HTML shell, the roast will tell you it found nothing readable — which is itself a finding, because that is also roughly what a search crawler and an AI assistant see when they try to describe your product to someone.

The headline names the product, not the outcome. "AI-powered workflow automation platform" tells a visitor what category to file you under and nothing about whether their afternoon gets better. The fix is a sentence about their situation after using you.

No specificity anywhere. Fast, powerful, seamless, intuitive, enterprise-grade. These words are free, therefore worthless — your competitors use every one. A single concrete number ("nine boards", "20 seconds", "refund any time in 30 days") outperforms a paragraph of adjectives, because a number is a claim someone could check.

The page never says who it's not for. Pages that try to convert everyone convert nobody, because a visitor can't tell if the promise applies to them. Naming your non-customer is the fastest way to make your customer feel found.

The CTA asks for more than the page earned. "Book a demo" above the fold, before a stranger knows what you do, is asking for a calendar invite from someone who doesn't know your name yet. Match the ask to the trust you've built by that point on the page.

The number is a rough band, not a benchmark, and we'd rather you use the three fixes than the number.

Below 40, the page has a clarity problem, not a design problem — a visitor can't tell what you do. Redesigning it will not help. Rewrite the headline first.

40 to 70 is the normal range for a page written by a founder in a hurry. It communicates, imperfectly. The listed fixes are worth real conversion.

Above 70, the page is doing its job, and your bottleneck is upstream — traffic quality — or downstream, in the signup flow. Go measure there before touching this page again.

Do them in the order given. The fixes are ranked by leverage, and the first one is usually the headline, which is twenty minutes of work and the largest single lever on the page.

Change one thing at a time and give it enough traffic to mean something. Changing five things at once and watching the number move teaches you nothing about which one moved it, which means you can't do it again on purpose.

Re-run the roast after each change. The score moving is weak evidence — it's a model's read, not a conversion measurement — but the fixes changing is strong evidence, because it means the previous problem is genuinely gone and something else is now the bottleneck. When the list stops changing, you've extracted what this tool can give you and the remaining work is measurement on real traffic.

One caveat worth stating plainly: a page can score well and still convert badly, because clarity is necessary and not sufficient. If your headline is sharp, your price is visible, your proof is real, and strangers still leave, the problem is upstream in who you're bringing to the page — and no amount of rewriting fixes the wrong audience.

If the page needs more than editing, kitty.build generates the whole thing from a validated idea — headline, structure, capture form — and publishes it to your own Cloudflare account on your own domain. Every revision is versioned, so a rewrite that loses conversion is one click to undo.