AI Slop Has a Look. Here's How to Spot It.

MUKUND PAREKH — Thu, 16 Jul 2026 09:31:24 +0000

A week ago I argued that the next competitive advantage won't be AI, it'll be taste. But I left one thing unanswered: if taste is the advantage, then what does the lack of it actually look like?

That question stuck with me. Because in the AI era, bad taste has a very specific look. Once you see it, you cannot unsee it.

Let me describe it.

The slop starter pack

You have seen this site. Probably five times today.

A purple-to-indigo gradient in the hero. A headline with "seamless" or "elevate" in it. Three feature cards, evenly spaced, each with a thin line icon. A bento grid underneath. A pricing table nobody asked for. And somewhere, if you look closely, a leftover "[Your Company]" that never got filled in.

It was built in an afternoon. It looks like it was built in an afternoon. Not because AI is bad, but because most people ship the first thing the model hands them.

That is the real tell. Not the AI. The not editing the AI.

The actual signals

Once I started paying attention, the patterns got specific:

Builder fingerprints. v0, Lovable, Bolt, Framer and friends leave traces in the markup: default class names, telltale script tags, untouched component libraries.
Untouched defaults. shadcn/ui and Radix straight out of the box. Great libraries. Dead giveaway when nothing is customized.
Copy tells. Lorem ipsum in production. Buzzword density. Em dashes everywhere, because that is how the models punctuate. (Yes, I notice. No, there are none in this post.)
Layout tropes. The gradient. The bento grid. The three cards. A "trusted by" logo row on a site with no customers yet.
Leftovers. "As an AI language model." Placeholder brackets. Alt text that describes the prompt instead of the image.

None of these prove a human was absent. Plenty of good sites use these tools. But stack five or six of them with zero editing, and you are looking at slop.

Taste is just noticing

Here is the thing about taste: it is not about avoiding AI. I use Claude and Cursor every day. Taste is noticing when your output looks like everyone else's output, and doing the extra 10% that makes it yours. The custom font. Copy that sounds like a person. The one odd detail a template would never include.

Speed is a commodity now. Taste is what is left.

So I built a mirror

I wanted to see this at scale, so I built Slopdar. You paste any URL. It reads the public HTML, runs about 50 of these checks,takes a screenshot, and gives the site a "Slop Score" from 0 to 100. Zero means it looks hand-crafted. 100 means prompt-and-deploy.

It is meant to be fun, not a verdict. It reports signals, not proof. A high score means a site smells templated, not that no human ever touched it.

Run your own site through it. It is a little uncomfortable and weirdly useful. If you want the full list of tells with how to check each one by hand, I wrote a guide here.

AI made building fast. It also made everything look the same.few years are the ones who can tell the difference, and whocare enough to fix it before they ship.

That is taste. And now there is a score for it.

The Next Competitive Advantage Won't Be AI. It'll Be Taste.

MUKUND PAREKH — Thu, 09 Jul 2026 11:11:09 +0000

When ChatGPT became mainstream, the biggest question was whether AI could build software.

Today, that question feels outdated.

With tools like Claude, Cursor, Lovable, Bolt, and GitHub Copilot, building software has become dramatically faster. A solo developer can ship in days what once took a small team weeks. The barrier to creating products has never been lower, and that's an incredible thing for our industry.

But lowering the barrier has also created a new problem.

As I browse Product Hunt, X, and Indie Hackers, I keep noticing the same pattern. Many new products look surprisingly similar. They have the same layouts, the same feature sections, the same gradients, the same testimonials, and often the same marketing copy with only a few words changed.

This isn't because AI is bad at building websites.

It's because AI is exceptionally good at producing the average first draft.

For a long time, execution was the biggest challenge. If you could build faster than everyone else, you had an advantage. AI has changed that equation. Speed is becoming a commodity. Thousands of developers now have access to the same models and can generate a polished landing page in a matter of hours.

That means the advantage is shifting somewhere else.

I think the next competitive advantage is taste.

Taste is knowing when a headline sounds generic. It's recognizing that a feature section doesn't communicate value. It's removing three sections instead of adding three more. It's making design decisions that aren't the obvious default. It's understanding your users well enough to know when the AI's suggestion is good, and when it completely misses the point.

AI can generate options.

It still can't replace judgment.

That's why I don't think we'll care much about whether a product was built with AI in the next few years. We'll care about whether the final product feels thoughtful, original, and genuinely useful. Users don't reward effort. They reward quality.

As developers, we're entering an interesting phase. AI is no longer the differentiator. It's becoming part of the standard toolkit, just like Git, Docker, or modern frameworks.

The developers who stand out won't necessarily be the ones using the most AI. They'll be the ones with the best judgment about what to keep, what to throw away, and what deserves to be built differently.