If you’ve noticed that websites have started to converge into one beige, sans-serif haze, you’re not imagining it.
One of AI’s biggest boons has been allowing non-technical folks to vibe code their ideas into real, monetizable apps. As we wrote in April, anyone can build an app in a couple of hours, using tools such as Claude Code, Lovable, Replit, or Base44.
These AI-designed apps have some telltale signs and the devil is in the details: using similar design styles that look pretty but are dysfunctional. While the apps may work at a small scale, these small details could become big problems when you scale up and go commercial.
Here’s how to tell if your app looks AI-coded and how to change it.
1. Regression to the mean, aka, painfully mid
The first sign: The app’s design is boring and cookie-cutter.
Paul Bakaus, the CEO of AI design startup Impeccable, said in a June 23 podcast interview with VC firm Andreessen Horowitz that AI giveaways — particularly for Claude Design — include beige or tinted backgrounds and sans-serif fonts.
He called it an “algorithmic Uniqlo or Ikea,” a design that’s not bad, but not necessarily unique.
Donghoon Shin, a human-computer interaction researcher at the University of Washington, published a paper about how vibe coding has led to the homogenization of design.
Shin told Business Insider that vibe-coded products tend to converge toward “a single, statistically average aesthetic.”
The hallmarks: a muted color palette with lots of whites and grays, a single brand accent color, standard sans-serif typography, and elements with rounded corners and drop shadows.
Screenshot/Aditi Bharade
Sauvik Das, an associate professor at Carnegie Mellon University’s Human-Computer Interaction Institute, told Business Insider that this was the “regression to the mean” effect.
People trying to make money off their AI-coded apps are noticing this too.
Priyanshi Bansal, a product manager in India, vibe-coded an app using Claude that helps people pick gifts for their loved ones. Early users told her the app looked like “AI slop.”
She said the first version of her app featured many emojis, shadows, and rounded edges.
“Now I’m building version two of the app, which is much better — particularly the UI, since I’m a product designer, so I need to nail that,” she said.
2. Pretty and dysfunctional
BI
The second telltale sign: The website is super posh for a product that’s not quite there yet.
“You’ll see, for example, a very polished landing page for a product that is still very much in an early alpha phase,” said Das.
It might look pretty, but a lot of the functions don’t work intuitively, because AI designs for aesthetics and skimps out on the usability.
User interface and experience designers are trained to read human emotions, behaviors, and intentions, said Ankush Samant, a lecturer of digital innovation and design practice at the National University of Singapore. They know exactly how to design “the weight of a button, the pacing of an onboarding flow, the tone of an error message.”
“AI tools optimize for the happy path and tend to produce interfaces that look complete until someone actually uses them,” Samant said.
Das added that vibe-coded products sometimes invite users to interact, even when that piece does not have a function. For example, hovering over an element gives it a subtly shaded outline and a size increase that suggests it can be clicked for more info, but clicking it yields nothing.
3. Error 404: Ignores edge cases
And lastly, vibe coding tools don’t spare much thought for edge cases.
“Designers also spend significant time on what I’d call edge-state design — empty states, error messages, skeleton loaders, and offline states — which AI often leaves as afterthoughts or omits entirely,” Shin said.
Samant said that how an app presents errors is how you can tell whether the app’s creator has thought through the full user experience, or if it is still in a demo phase.
“AI tools tend to either skip these entirely or generate placeholder copy — ‘Something went wrong. Please try again’ — that strips the human voice out exactly when users most need reassurance,” he said.
To be sure, vibe coding startups are taking note of all these drawbacks.
San Francisco-based startup Base44 on Monday launched its own AI model, Base 1, which it hopes will reduce the AI-slop look of vibe-coded products and produce better UI/UX than the frontier models.
You vibe-coded an app. Now what?
In this house, we don’t shame vibe coders. We’re big enthusiasts ourselves.
But if you find that your app has some of the red flags mentioned above, here are some ways to change them.
Samant said the first step is to stop prompting for aesthetics and start prompting for decisions.
“Instead of ‘make this look clean and modern,’ try: ‘This screen is where a user who is anxious about their data will decide whether to continue. What should I remove, and what does the copy need to do?'” he said.
Secondly, vibe coding isn’t all about the vibes. Shin said users should provide as many specifics as possible to AI tools, such as design references, brand constraints, and what the user doesn’t want to see.
And if you’re thinking of scaling up, there is a case for hiring a professional to take the product to the next level.
Das said that while AI is a helpful tool for improving user interfaces, it might not be the best for creating good UIs from scratch.
“I won’t say that a purely vibe-coded app cannot achieve sustained commercial success, because the most important thing there is product-market fit,” he said. “But I think a good UI/UX designer is always necessary to take a good product and make it great.”
