Loading
The web scan is free and needs no account.
Up to 25 pages of your site, checked for the eight defaults AI builders leave behind, page by page. Then one fix prompt per default, naming the pages and the classes, ready for Claude Code, Codex, Cursor or GitHub Copilot, or the chat of Lovable, v0, Bolt, Replit.
Preview free. Full report 19 EUR, once.
19 €once
You are editing the code of your-site.com. A scan of 18 pages of the live site found "Default blue" on 14 of them: - / - /pricing - /blog - and 11 more Evidence on the pages: bg-blue-600, text-indigo-500 Goal: define your own accent and replace the family with it. What to do: 1. Search the source for the blue, indigo and sky utility families (bg-, text-, border-, ring-, from-, to-) ... 2. Replace each one with the accent token below ... Design tokens generated for your-site.com: - accent: (a colour picked for your domain) ... Literal replacements found on the pages read: bg-blue-600 -> bg-[var(--accent)] ...
Shortened, on a made-up domain: your report fills it with your own pages, the classes found on them and the tokens generated for your domain.
The free scan reads one page and gives that page a fix kit. The report reads up to 25 pages of the site, tells you which default sits on which page, and writes one prompt per default that names those pages, so an agent fixes the shared component once instead of you hunting page by page. It also adds a rules file to keep the defaults out of the next generated page, a Markdown download, and 3 reruns to check the fixes.
Use them, and use our free one-page scan: a score for one page and a generic prompt cost nothing anywhere now. The report is for the next step: 25 pages read, which default sits on which page, a prompt per default with the literal classes to change and the tokens to use instead, and the rules in the file each tool actually reads (AGENTS.md, CLAUDE.md, Cursor rules, builder instructions), so the next generated page does not bring the defaults back.
No. The pages are fetched and checked against eight signatures with fixed rules, and the prompts are assembled from what was found: the same site gives the same report. The prompts are written for your AI tool to apply, so the editing happens in your code, under your eyes.
You can, and with no evidence it guesses. The report gives the agent the facts it cannot see from a chat: which pages, which literal classes, how many of them, and a palette and type pair that stay the same across every page. That is the difference between "make it less generic" and a diff you can review.
People who built a site with Lovable, v0, Bolt, Replit, Cursor, Claude Code or Codex, or a starter template, and want it to stop looking like the next one, and studios checking a client site before handing it over.
Pages behind a login, pages that refuse automated requests, and anything not linked from the sitemap or the home page. On Tailwind sites most classes live in compiled CSS, which the report reads with the pages; your source can hold more than the rendered pages show.
By card through Stripe, 19 EUR once, VAT included. No account and no subscription: the report link is your access, and refunds are given within 14 days if it did not help.
The report itself, under a private link with a secret key in it, so you can come back to it. Unpaid previews are deleted after 30 days. The pages read are public pages; nothing is shared with anyone.