Free tool

AI slop detector: does your site look AI-generated?

Enter a website and see how many of the tell-tale AI design habits it has, from purple gradients and emoji icons to buzzword copy, with what to do instead.

By the UXKIN team · Reviewed October 5, 2026

Free, no account. Public pages only; nothing is saved.

What the detector looks for

AI website builders and coding agents share a handful of habits. Each one is harmless alone; together they make a page look like every other generated page. The detector reads the page and its stylesheets and checks for these:

  • Purple-to-blue gradients, in backgrounds, buttons and especially headline text.
  • Emoji as icons: 🚀 ✨ ⚡ in headings, feature lists and buttons.
  • Buzzword copy that could describe any product: “unlock the power”, “seamless”, “supercharge”, “to the next level”.
  • The three-card feature grid: icon, heading, two lines, three times.
  • Everything centred, including long text and cards.
  • Glowing blobs and frosted glass behind the content.
  • The same big rounded corners and soft shadows on every element.
  • Generic buttons like “Get started” and “Learn more”, vague social proof (“Trusted by 10,000+ teams”), placeholder content and only the default font.
  • Newer models' writing habits, which show up even on carefully built pages: the same promise repeated everywhere (“no account, no upload”), calling things “honest” or “genuinely” this and that, explaining everything by what it isn't (“rather than…”), a “who it's for” grid of job titles, “No X. No Y. Just Z.” lines and all-caps claims strung together with ✦.
  • App-builder habits (Lovable, Bolt and similar): “Stop X. Start Y.” slogans, testimonials signed “Marcus P.” saying “Life-changing!”, a permanent “Save 50%” against a higher crossed-out price, big round stats like “50,000+ users” and Every Heading In Title Case.

How the score works

Each habit has a weight: the ones people notice first, like gradients, buzzwords and emoji, count most. The score is how much of that weight the page shows, from 0% (looks hand-made) to 100% (full AI slop). It's deliberately strict: under 12% is a page with its own character; 12–34% has a few AI tells worth fixing; 35–59% looks noticeably generated; 60% and up is the full template.

It isn't an AI detector for who built the page. It measures the design habits, which is what visitors actually react to.

How to make a site look less AI-generated

Most fixes are small, and they share one idea: replace the defaults with decisions. Use your brand's own color instead of a gradient. Write what the product does instead of how amazing it is. Show the product instead of three icon cards. Left-align anything people read. Then look at how a few real products in your category handle the same sections, and borrow their decisions, not their pixels.

A prompt for your coding agent

Paste this into Claude Code, Cursor, Codex or Copilot along with your detector results:

Prompt
Our site looks AI-generated. Fix these without changing what the page says:
- Replace purple-to-blue gradients and gradient text with our brand color, mostly solid.
- Remove emoji used as icons; use one icon set, or none.
- Rewrite headings and buttons in concrete words: what the product does, for whom, and what each button does.
- Left-align sections and cards; centre only short hero text.
- Replace the three-card feature grid with one real screenshot or a short demo of the product.
- One corner radius for the whole page; shadows only on things that float.
Before changing anything, look at how 2 or 3 real products in our category handle their hero and pricing sections, and say which decisions you're adopting.

Questions

What is AI slop in web design?

It's the look that AI website builders and coding agents fall back on when nobody gives them direction: purple-to-blue gradients, glowing blobs, emoji as icons, three identical feature cards, everything centred and copy full of buzzwords that could describe any product. Any one of these is fine on its own; together they make a site look generic and generated.

Is the AI slop detector free?

Yes. It's free, needs no account and has no limit on checks. It checks public pages only and doesn't keep who checked what; a site's result is remembered for a few minutes, so checking it again is instant.

Does it tell me whether AI made my site?

No, and nothing can reliably. It measures how many of the common AI-generated design habits a page has. A human can build a page full of them, and a site made with AI can score low if someone gave the agent real direction.

My page shows a bad example on purpose. Will it count?

Not if you mark it. Add a data-slop-example attribute to the element that holds the example (a "before" mock-up, say) and the detector skips it and everything inside it.

Does it work on sites built with Lovable, Bolt, v0 or a React app?

Yes. Those sites build their pages in the browser, so the page itself arrives almost empty. When that happens the detector reads the app's code instead, where all its text and styles live, and checks every screen of the app at once. The result says when it did this.

How do I stop my coding agent from producing AI slop?

Give it evidence instead of adjectives: your own design system, a clear brief, and real examples of how good products solve the same screen. That's what UXKIN does: it gives Claude, Cursor, Codex and other agents real app screens, user flows and website design systems to work from.

Put it to work

UXKIN gives your coding agent real app screens, user flows and website design systems to work from, so what it builds looks like a real product, not the same AI template. Browsing the library is free.