When you build a website with an AI tool, you leave traces. Not intentionally — it is simply that these tools have characteristic ways of structuring code, loading assets, and configuring infrastructure that are recognisable once you know what to look for.
Our AI Website Detector identifies over 20 different AI builders. Here is exactly how detection works for the most common ones.
How Detection Actually Works
AI website detection uses several different signals simultaneously. No single signal is definitive — it is the combination that produces a reliable identification.
1. CDN and Hosting Fingerprints
Different AI builders use different hosting infrastructure. Framer sites are almost always served from Framer's CDN. Vercel v0 sites typically use Vercel's edge network. These CDN signatures appear in HTTP response headers, IP ranges and CNAME configurations — invisible to users but highly readable by detection tools.
2. CSS Class Names and Conventions
AI builders generate CSS using characteristic naming conventions. Framer uses class names like framer- prefixes throughout. Wix generates classes with _1 and _2 numeric suffixes. Bolt.new and Lovable produce utility-class patterns that differ from hand-written or traditional framework CSS.
3. HTML Comments and Generator Meta Tags
Many AI tools leave explicit breadcrumbs. The <meta name="generator"> tag is the most obvious — tools like Wix, Squarespace and older AI builders often include their name here directly. Others leave comments in the HTML source: "" or similar.
4. JavaScript Bundle Signatures
The JavaScript bundles produced by different frameworks and AI tools have characteristic sizes, chunk patterns and import structures. Framer's React-based output looks different from Webflow's generated JS, which looks different from a custom Next.js build. The bundle fingerprint is one of the most reliable signals.
5. API Reference Scanning
Some AI builders make direct API calls back to their own infrastructure — for analytics, form handling, CMS content, or real-time features. Finding a call to api.framer.com or cdn.builder.io in the network requests is a definitive identification.
The Most Detectable Builders
Framer: Extremely detectable — uses its own CDN, generates characteristic React component structure, and often includes Framer attribution in the page source.
Wix ADI: Highly detectable through Wix's proprietary hosting infrastructure, _wix_ prefixed resources, and static.wixstatic.com CDN.
Bolt.new / Lovable: Detectable through characteristic Vite build output, specific deployment patterns on Netlify or Vercel, and occasional HTML comments.
Webflow: Strongly detectable through webflow.io subdomains or webflow.com scripts, even when deployed to custom domains.
The Harder Cases
Some AI-assisted builds are genuinely difficult to detect. A developer who takes a v0 or Copilot-generated component, heavily modifies it, and deploys it on their own infrastructure may leave no detectable traces at all. Detection works on the output — not the process. If the output looks like hand-written code on custom infrastructure, no tool can reliably identify it as AI-assisted.
Why Does This Matter?
There are several practical reasons you might want to know whether a site was built with AI tools: evaluating a developer's claimed skills, understanding a competitor's tech stack, or assessing whether a site has the custom engineering its owners claim it does.
Try our AI Website Detector on any URL — it checks all the signals above and tells you which AI builder was used, if any.
