What is vibe coding?
Vibe coding is building software by describing what you want to an AI and shipping whatever comes out: coding by vibes instead of by hand. You prompt, the model writes, you look at the result instead of the code, and you iterate by asking for changes. It's the fastest path from idea to deployed website that has ever existed, which is exactly why the web is now full of it.
Where the term came from
Andrej Karpathy coined it in February 2025, describing a way of working where you "fully give in to the vibes" and forget the code even exists. He meant it half as a joke about his weekend projects. The internet adopted it as a job description. Within months, tools like v0, Lovable and Bolt had turned vibe coding from a workflow into a product category: type a sentence, get a deployed website.
AI-assisted development
A developer uses AI to move faster but stays in charge: reads the generated code, rewrites what's wrong, makes the design decisions, owns the result. The AI is a power tool. The output usually doesn't look generated, because the fingerprints got sanded off.
Vibe coding
You describe what you want, accept what comes out, and iterate by prompting rather than by reading code. Great for prototypes, demos and scratching an itch. The trade: you ship the tool's defaults, and the tool's defaults look the same on every site it produces.
Prompt-and-deploy slop
Vibe coding minus the taste: generate, deploy, walk away. Nobody read the copy, nobody replaced the placeholder text, nobody asked why the hero gradient is indigo-to-violet again. This is where vibe coding turns into slop.
Why vibe-coded sites all look the same
A language model asked for "a modern landing page" produces the statistical average of every landing page it trained on: gradient hero, three feature cards, bento grid, indigo-to-violet palette, Inter font, shadcn/ui components. None of these are wrong. All of them together, untouched, means no human made a single visual decision. That sameness is what makes vibe-coded sites detectable, and it's the difference between using AI and publishing slop.
How to spot a vibe-coded website
The tells are consistent: builder fingerprints in the source code (the tools literally sign their work; here's how to check for v0, Lovable and Bolt specifically), untouched default stacks, AI writing patterns, leftover placeholders and default metadata. We keep the full ten-point checklist in how to tell if a website is AI-generated, or you can scan all of them at once.
Vibe coding isn't the crime. Shipping the defaults is.
A vibe-coded prototype that later got real copy, real design decisions and a proofread will score low on any honest detector, because the fingerprints are gone. That's not evasion; that's craftsmanship. Slopdar measures the fingerprints left behind, not whether AI was used. Here's exactly how the scoring works.