How to Tell if a Website Was Built with Lovable

Updated October 15, 2026. 6 min read

Lovable is one of the most popular AI app builders: you describe a website in a chat, and it writes and publishes the code for you. Sites built this way often look polished, so it isn't always obvious that nobody hand-coded them. Here is how to check, from a ten-second glance to a look at the source.

A quick note on fairness before we start. A site built with Lovable is not automatically worse or less honest than any other site. Plenty of real businesses use it. What usually matters is whether the content and the people behind it are real. Knowing how a site was made is just one useful piece of that picture.

The ten-second checks

The "Edit with Lovable" badge

Projects published on Lovable's free plan show a small "Edit with Lovable" badge on the live site, usually in a bottom corner. Paid plans can hide it with a single setting, according to Lovable's own documentation, so a missing badge proves nothing. A visible one settles the question.

A lovable.app address

Every published Lovable project gets a free address ending in .lovable.app. If the site you are looking at lives on such an address, it was built and hosted with Lovable. Sites on their own domain can still be Lovable projects, so keep going.

Checking the source code

Open the page source with Ctrl+U (Cmd+Option+U on a Mac) and use the browser's search to look for these strings.

lovable-uploads

Images that a site owner uploads inside the Lovable editor typically end up under a path containing /lovable-uploads/. This is one of the most dependable signs we have seen, because the path survives when the project moves to a custom domain and nobody thinks to rename it.

data-lov-id attributes

Lovable can tag page elements with attributes named data-lov-id so its editor knows which component you clicked on. They are meant for editing, but in some published projects they remain in the live HTML. Real websites have no other reason to carry an attribute with that name.

Leftovers from the GPT Engineer days

Before it was called Lovable, the product was known as GPT Engineer. Older projects sometimes still load a script from cdn.gpteng.co, for example gptengineer.js. If you see that domain in a script tag, the site started life in the tool.

Clues from the technology stack

Lovable generates code with a fixed set of building blocks, so the stack itself is a weaker but useful hint.

• Tailwind CSS with shadcn/ui. Look for class names such as bg-background, text-muted-foreground and border-input. These are the default tokens of the shadcn/ui component kit, which Lovable and similar tools reach for out of the box.
• Lucide icons. Inline SVG icons with classes like lucide lucide-arrow-right.
• A Supabase backend. Open the browser's developer tools, switch to the Network tab and reload. Requests going to an address ending in .supabase.co mean the site uses Supabase, one of the backends Lovable connects to.

Every one of these is also popular with human developers. Treat them as supporting evidence that makes a stronger sign more believable, never as proof on their own.

The empty page test no longer works for new projects

For a long time, Lovable sites shipped an almost empty HTML file: a single <div id="root"> plus a script named like /assets/index-a1b2c3.js, with all the text filled in by JavaScript. That pattern is still common on older projects. However, Lovable's documentation says that apps created from May 13, 2026 use TanStack Start and render pages on the server, so newer Lovable sites deliver full HTML like any traditional website. If you only rely on the empty-page test, you will miss them.

What does not prove a site was built with Lovable

We got this wrong ourselves in an early version of Bot or Built, so it is worth spelling out.

• A link to Lovable. A company that lists Lovable as a customer, partner or tool it likes will have links to lovable.dev on its pages. Stripe's homepage, for example, shows Lovable in a customer carousel. That says nothing about how Stripe's own site was built.
• A link to a separate Lovable page. We came across a company whose main website was built conventionally, while one promotional page it linked to was hosted on lovable.app. Technology trackers had listed the whole company as a Lovable site because of that single link.
• The word "lovable" in the text. It is an ordinary English word, after all.

The rule we follow: only count a match when it shows up in the site's own code, such as an image path, a script, a stylesheet or an attribute, and not in visible text or an ordinary link.

Checking automatically

Bot or Built looks for all of the code-level signs above, plus fingerprints of other AI builders, and lists every clue it finds so you can judge them yourself. A confirmed Lovable fingerprint puts a site at "Mostly bot" or higher, because it means the site itself was generated. It doesn't tell you whether the text on it was written by a person, so read the content with the same care you would give any other site.