Is My Website AI Slop? A 20-Minute Self-Check for Builders

Is my website AI slop? It is when a first-time visitor can't tell what you do, can't tell your site from every other site, or can't finish the one thing you want them to do. You can test all 3 in 20 minutes with your live site and your phone, asking one question for each of the six areas our audit scores.

A first draft built with AI often ships before anyone has used it the way a visitor would. This guide gives you a timed routine for doing that, and a marked-up example of what finished notes look like.

What are we checking for?

Not whether AI touched your site. Whether you built in Lovable, Bolt, Replit, or another tool, a visitor only asks whether it is clear, believable, and working. If you want the definition and history of the word first, start with what AI slop means.

So the routine has one question for each of the six areas our audit scores: Clarity and Promise, Craft and Design Integrity, Trust and Legitimacy, Flow and Usability, Mobile and Access, and Conversion and The Ask. You are doing the looking by hand here, so write down what you see, not what you meant.

How do you set up the 20 minutes?

Spend the first minute on setup.

1. Open your live site at its public address, not the preview inside your builder.
2. Use a private window, so you are logged out like a new visitor.
3. Open the same address on your phone.
4. Start a notes page with 3 headings: What works, Slop tells, Top fixes. Our report uses the same 3 fields for each area, and the example near the end shows them filled in.
5. Set a timer for each check, and stop when it rings.

Does the first screen say what you do?

3 minutes. This is Clarity and Promise. Look only at the first screen, then read the headline aloud and try to finish this sentence from it alone: "This is a _ for _ who want to ___." Then check that the rest of the page agrees with it.

Vague lines about unlocking possibilities fail because they fit any site. A line passes when a visitor could repeat it back in one breath. If you can't finish the sentence, write the gap under Slop tells, and write the line you wish it said under Top fixes.

Does it look made for you, or made for everyone?

4 minutes. This is Craft and Design Integrity. Scroll the whole page slowly and look for places where nobody made a choice:

• Generic art or colored blocks where a real screenshot, product photo, or face should be.
• One style of icon repeated beside every feature.
• Rows of identical cards, with every section laid out the same way.
• Buttons, corners, and fonts that change from one section to the next.
• Text you could paste onto a competitor's site unchanged.

Then ask what on this page only you could say. If the answer is "nothing," that is the biggest item for your Slop tells list. One small extra: if an image carries meaning, such as a chart or a product shot, it needs a text description, while a purely decorative one doesn't (W3C alt decision tree, page updated 13 May 2024, read 5 October 2026).

Does the site give a visitor a reason to trust it?

3 minutes. This is Trust and Legitimacy. Put yourself in the seat of a cautious customer who is about to hand over money or personal details, and ask who is on the other side:

• Is there a named person or business, and a way to reach them?
• Does anything on the page suggest the site is still looked after, or does a date look frozen?
• Do the privacy and terms pages open, and do they describe this product rather than a template?
• Is there any leftover template text, like "Your Company Name" or "Lorem ipsum"?
• Can you trace each testimonial to a real person and each number to a source?

Checking that legal pages exist and match what you do is a good habit, but nothing here is legal advice. A trust signal only counts when it is true, so don't add a counter or testimonial unless the number or person is real.

Does the main path work when something goes wrong?

4 minutes. This is Flow and Usability. Pick the one thing a visitor is meant to do (sign up, book, buy, or send a message) and do it start to finish. Then break it on purpose:

• Submit with a field left empty.
• Type a bad email address.
• Press back midway, then try again.
• Click every menu and footer link.

Does each error say what to fix? Does the site confirm that it worked? Does the confirmation email arrive? If you submit a form, use a test email you control and label the entry "test," and stop before any real payment. Anything you couldn't check goes in your notes as unconfirmed, which is more honest than guessing. The full workflow is in our guide on how to test a vibe-coded app.

Can someone use it on a phone?

3 minutes. This is Mobile and Access. Take your phone and open the live site fresh, so any pop-ups or notices appear the way a new visitor would meet them. Check 4 things:

1. On first load, can you spot the main button right away, with nothing sitting on top of it?
2. Can you read the text without zooming?
3. Does anything run off the side of the screen?
4. Can you tap each menu item without hitting its neighbor?

Do this on a real phone, not a shrunken desktop window. Pop-ups, fixed bars, and chat widgets can all cover a button on a small screen.

Is it clear what you want visitors to do next?

2 minutes. This is Conversion and The Ask. Find the main button and read only its label. Could someone who skipped the page say what happens when they click? Is there one main action, or 5 competing ones? Is the price, or the fact that it is free, visible before the ask?

A label that names the action and the cost tells a visitor what they are agreeing to. A vague label makes them guess, and guessing is a reason to leave.

How do you tell if a website is AI generated, or just AI slop?

We can't tell reliably by eye, and a visitor doesn't need to. Some AI-built sites are thoughtful, and some hand-built ones are generic. A vague headline or a row of identical cards is a hint, not a verdict.

"AI generated" is about how a site was made. "Slop" is about whether anyone checked the result. So swap the question for 3 smaller ones:

1. Could the headline fit any site in your category?
2. Can you name 1 choice on the page that someone made on purpose?
3. Did anyone run the main path on a phone after the last change?

3 honest yeses are a good sign that someone cared for the site, whatever tool built it. Any "no" belongs in your Slop tells list.

How do you write up your notes and decide what to fix first?

Mark each area Pass, Maybe, or Fail, with one line under each heading. Then pick the order: start with the answer that made you wince most. A broken main path usually earns that spot, because nothing else matters if a visitor can't finish.

Hypothetical example: a made-up resume-feedback app, after the 20 minutes.

Under a last heading, Not confirmed, the writer adds: "Feedback email: nothing arrived in 10 minutes, spam folder checked." They circle the upload error, since it made them wince most, and list the fixes in order: file types first, then the price, then the subheadline. That is the shape to aim for: a short, ranked list, not a rewrite.

If you want to see how a finished report lays this out, the sample report uses fields such as Verdict, What works, Slop tells, Top fixes, Where to start, Recorded evidence, and Not confirmed.

What should you do after the check?

Fix the top 3 items, then run the same 20 minutes again. You know your own site too well, so ask one other person to run the main path while you watch without helping. That people round comes after you have fixed the obvious problems.

If you would rather have an outside read, Is My App Slop looks at the same six areas and ranks the fixes. Our guide to AI slop detectors explains what scanning tools can and cannot see. Once your list is short, our guide on shipping or iterating can help you decide when it's ready to share.