You’ve seen this website before. Not this exact one, but its shape: a soft purple gradient up top, three identical feature cards in a row, a cheerful button that says “Let’s Go!” It looks sharp, it loads fine, and it feels like a hundred other pages you scrolled past this week.
That familiar feeling has a name. People call it AI slop, the fingerprint AI leaves behind, and once you can spot it you start seeing it everywhere: in apps prompted into existence overnight, in sites spun up in an afternoon, in articles that hit every point and land none of them.
Welcome to the machine. It’s meant as a warning about an industry that swallows people whole, and plenty of workers feel that same dread about AI right now: that it’s coming for their work and their livelihoods, that CEOs are planning to lay off entire departments and staff them with a few senior staff and an army of machines. It’s a legitimate concern, but it points in the wrong direction.
A tell is not a verdict. Spotting AI in an app, a site, or a piece of writing doesn’t mean the thing is broken or worthless. It means a machine did the work and nobody came back to make it theirs, which is sometimes harmless and sometimes a real liability hiding under a polished surface.
Being able to recognize where AI started a project that a human never finished is an important skill to have now that these tools are everywhere.
This two-part guide walks through both halves of that skill. Part 1 stays on the surface: how to spot AI in an app, a site, or a piece of writing, and how to read what you find. Part 2 goes underneath an AI-built app, into what it does with your data, the questions that tell you whether it’s safe, and how a shaky one gets fixed.
What Vibe Coding Actually Is
A few years ago, building an app meant knowing how to code. Now it can mean knowing how to ask. Tools like Lovable, Bolt, and v0 let you describe what you want in plain English. Tell AI that you want “a booking page for my salon, with a calendar and a way to reach me,” and watch a working version appear in seconds. That’s vibe coding: you bring the idea, AI writes the code.
For getting off the ground, these tools are a real head start. Setting up a new project by hand can eat half a day before you’ve built anything a customer would notice. A prompt does that same thing in minutes, which for someone with an idea and no development background is the difference between starting and never starting at all.
A working version isn’t a finished one. What you get back looks the part and usually runs fine in a quick demo, but it’s where you start, not where you stop. It’s the first rough draft of your front door, assembled from the most common patterns AI has seen, and it still needs a person to secure it and make it truly theirs.
That distance, between “it runs” and “it’s ready,” is what the rest of this post will help you spot.
What an AI-Built Site or App Looks Like
The giveaways are visual. Once you’ve seen them, you’ll recognize them again. AI builds from the most common patterns it was trained on, so it reaches for the same safe defaults every time, and those defaults show up together across sites and apps that otherwise have nothing in common.
Here’s what to look for.
- A purple-to-blue gradient. This is the most reliable one. It shows up on the hero background, the buttons, even the little icons, because it’s the default accent baked into the tools these apps are built with. Once you notice it, you’ll see it on landing pages everywhere.
- Everything matches a little too perfectly. Identical spacing, identical rounded corners, one typeface doing every job, three feature cards of exactly equal size. Real design uses variation to guide your eye, and AI tends to make everything uniform, which reads as flat.
- A thin colored bar down the left edge of cards and boxes. A slim vertical stripe, usually in that same purple or blue, running down the side of quote blocks, callouts, and info panels. It reads as a deliberate design choice, but it’s a default the tools lean on, so the exact same stripe turns up on sites that have nothing to do with each other.
- Soft, glowy, and rounded past the point of purpose. Buttons with a gentle glow, drop shadows on everything, corners rounded into pill shapes, and the odd emoji dropped in for decoration. Any one of them is harmless. All at once, they’re a look.
- Buttons and messages that sound like a pep talk. A hero that offers to “supercharge” or “unlock” something, a call to action that just says “Get Started,” a success note chirping “You’re all set!” It’s the voice AI defaults to when no one has told it how the business actually talks.
- Nothing moves the way you’d expect. You hover over a button and it does nothing, or the whole page fades in with the same generic animation on every element. Purposeful motion takes a human deciding what deserves attention, and that step usually gets skipped.
- Imagery that feels a half-step off. Illustrations that are a little too smooth and symmetrical, with a faint plastic sheen, or stock photos of a “team” that isn’t yours standing in for the real thing.
None of this is hard to change, which is why it isn’t the real story. A studio can swap in a brand color, a real typeface, and a photo of your actual storefront in an afternoon, and the AI look lifts. It only ever told you a machine got it started and no one came back to finish the styling; what’s underneath is a different question entirely.
What AI-Written Text Reads Like
The look is one giveaway; the words are the other. AI-written text is easy to read and hard to remember. The grammar is flawless and every point is covered, but the sound of a specific person who has actually experienced what they’re writing about is missing.
Once you tune your ear to that absence, you start to hear the same handful of habits.
- The same stock words, over and over. AI reaches for a small, predictable set: delve, leverage, robust, seamless, elevate, unlock, landscape, testament. One of them is nothing. A paragraph wearing all of them is a fingerprint.
- Everything sounds equally important and equally smooth. Sentences that all run about the same length, paragraphs that all run three or four lines, transitions so even that nothing ever catches you off guard. Real writing has bumps; it speeds up, slows down, and lingers where the writer actually cares.
- Lists that always come in threes. A single trio is good rhythm. But when every list in a piece is a neat set of three, it starts to feel like the writer set a timer. Human writers use it too. The first two paragraphs of this post stack three of them in a row, one of which is about three identical cards, so we’re not throwing stones from a clean house.
- The “not this, but that” setup, on repeat. AI likes to tee up one idea just to knock it down: it’s not X, it’s Y. Once, it’s a sharp way to land a point. Every few sentences, it’s a rhythm no person actually talks in.
- Sentences that open on a dramatic “Because.” AI makes a claim, then starts the next sentence with “Because,” dangling it like a reveal. It forces weight onto a point instead of earning it.
- Dashes dropped in everywhere. The em dash, that long dash marking a dramatic pause in the middle of a sentence, turns up in AI writing far more often than most people would use it. Most writers use a regular hyphen for nearly everything and rarely reach for the longer dashes at all. One here and there is normal. A page where every other sentence has one is a signal.
- Every section ends by repeating itself. AI likes to close each part with a tidy sentence restating what it just said, a little bow on top. It reads as thorough and adds nothing new.
- Polished, and completely hollow. Here’s a test you can run in ten seconds: drop a competitor’s name into the copy where the business name sits. If it reads just as well with their name on it, the writing is too generic to belong to anyone, which is exactly what AI produces when no one gives it something real to say.
A generic paragraph never lost anyone their savings, and neither did a purple gradient. Both tell you the same thing, that a person didn’t stay to make it theirs, and neither tells you whether it’s safe to use. That last question is what actually costs money.
This is Part 1 of a two-part guide. Everything here has been about the surface, the part you can see, and the part that can’t really hurt you. Continue to Part 2 for the part that can: what’s underneath an AI-built app, and how to tell whether the one in front of you is safe to trust.