Last updated
How to Tell if Your Designer Used AI Slop: A Founder's Checklist
Author
You paid for design. The screens look clean, the gradient is nice, and your investor deck even got a compliment. But something feels amiss. Your product looks just like the competitors you saw on Product Hunt last week, and once you click past the demo flow, things start to fall apart.
Trust that feeling. It's usually the first sign you've got AI output with a human invoice.
Let's be clear: your designer using AI isn't the problem. We use AI every day at Foundey, and so does every good designer we know. The real problem is AI output that nobody reviewed or shaped for your product. That's AI slop. It's not about crooked buttons or clashing colors. It's design with no author.
Most founders aren't designers, and most guides on spotting AI design are written for other designers. This one's for you. Here are five tests you can run in 30 minutes, no design background needed. Plus, a scorecard and a plan for next steps.
The short answer
To spot AI slop, look for decisions nobody made. Are the fonts, colors, and layouts just tool defaults? Click through void, error, and mobile states. Open the Figma file; does it have real components and version history? Ask your designer to explain three specific choices. One sign alone means little. A pattern across all five tests usually tells the story.
AI-assisted vs. AI slop: the line that matters
Before you start, get clear on the difference. It changes what you look for.
AI-assisted Design | AI Slop |
Starts from your users, positioning, and constraints | Starts from a prompt like "modern SaaS dashboard" |
AI generates options; a designer chooses, edits, and rebuilds | The first decent output ships with light tweaks |
Built on your design system and tokens | Built on tool defaults: Inter, the Tailwind palette, stock card components |
Unhappy paths are designed on purpose | Only the demo path exists |
The designer can explain every major choice | "It looks clean" is the whole rationale |
Notice what's missing: speed. A senior designer using AI well can deliver in two days what used to take two weeks. If the work passes these tests, fast delivery is a win. If you want the deep dive on AI slop and why it happens, we've covered that elsewhere. This guide is about judging the work in front of you.
Test 1: The five-minute screen scan
Open your main screens side by side: homepage, signup, dashboard, settings. Look for defaults. One is fine. Several together means nobody was making choices.
Typography nobody chose
Inter is a great typeface, which is why it's the default in almost every AI builder and component library. Seeing it isn't a red flag. Seeing it everywhere, with no clear difference between headings and body text, and random semibold weights, is. Ask yourself: if you covered the logo, would the type say anything about your company?
Color as decoration
Purple-to-blue gradient hero. Subtle glow behind dark sections. Different colored stripes on every card. The question isn't if they're pretty. It's if each color means something. In good work, one color always means 'primary action,' another means 'danger,' and you can predict which is which even on a new screen. In slop, color is just wallpaper.
Everything the same size, shape, and weight
Identical cards, same corner radius, same padding. Three-column feature grid. Bento box layout because it's trendy. When everything has equal visual weight, nothing stands out. Users have to figure out what matters. Hierarchy means some things are clearly more important.
Copy that could belong to anyone
"Supercharge your workflow." "The all-in-one platform for modern teams." Placeholder names such as Acme Corp or Joe Bloggs. Demo numbers that are too round. Copy is often a louder tell than visuals. Quick check: if your two closest competitors could use your headline without changing a word, the copy wasn't written for you.
Icons, emoticons, and dots that mean nothing
Emojis as navigation icons or section titles. Rockets and sparkles next to features. Colored status dots on items with no status. These are just decorations masquerading as information, and they train users to ignore the signals that matter.
Test 2: Click past the happy path
This is the test that matters most. This is where AI slop starts costing you money. AI tools nail the demo screen. They're weak everywhere else, because in the demo, the account always has data and nothing ever fails.
Empty, error, and loading states
Create a new account. What do you see before there's any data? A cheerful illustration and 'Nothing here yet' is just a placeholder. A real empty state tells a new user exactly what to do first. It's one of the highest-leverage onboarding moves in any SaaS product.
Now break things on purpose. Submit a form with an error in the email. Turn off Wi-Fi and click save. Upload the wrong file type. A good error message tells you what happened and how to fix it, in plain language. A slop error just says 'Something went wrong' in a red box.
In our audits, empty states and error handling are usually the first places AI-built products fall apart. You won't see them in a portfolio, but they're critical for activation.
The same action, five different ways
Find the primary button on five screens. Is it the same color, size, and label style each time? Now check destructive actions: delete, remove a teammate, cancel a subscription. Do they look and behave the same? AI-generated interfaces are often consistent on each screen, but inconsistent across the product. Each screen looks fine on its own because it was generated separately.
Everything on screen at once
A data table, input form, and export panel all on one screen. A settings page with 40 toggles and no grouping. AI tends to show every option at once because it doesn't know what the user needs right now. Good design only shows controls when needed. When we reworked the GoAudience dashboard, we focused on transparency and speed for marketing teams, which meant deciding what to leave out at each step.
The phone test
Open the product on your real phone using cellular data. Do buttons fit your thumb? Does navigation still make sense? Do tables turn into unreadable horizontal scrolls? AI tools make layouts that technically reflow on small screens, but it's obvious nobody actually used them there.
Test 3: Open the Figma file
You paid for design, so you should have the source files. If you don't, that's your first red flag. Once you're in, you don't need design skills to spot the signals:
Components. Click a button. If the right-hand panel shows it's an instance of a main component, the same button is defined once and reused. If every button is a separate loose shape, there's no system.
Named styles or variables. Colors and text styles should be named based on purpose, like 'action/primary' or 'text/secondary' not just raw hex codes everywhere.
Page structure. Look for pages that separate flows, states, and archived explorations. One giant canvas called 'Page 1' is a warning sign.
Layer names. Hundreds of layers called 'Frame 1847' or 'Rectangle 12' usually mean the work was generated or rushed.
Version history. Open it up. Real design shows weeks of iteration, exploration, and rejected directions. A few big drops with nothing in between means screens were made elsewhere and pasted in.
States. Are the empty, error, and loading screens from Test 2 actually designed in the file?
One note: some strong designers now work directly in code. If that's your setup, ask for the equivalent: the token file, the component library in Storybook, or the repo. The question is the same: is there a system, or just a pile of screens? A system lets you change your brand color in one place and keeps future AI-generated work on brand. We wrote a full guide on building a design system that survives AI tools.
Test 4: Ask five questions
The best AI slop detector isn't software. It's a conversation. People who made real decisions can explain them. People who just accepted defaults can't.
Question | A strong answer sounds like | A red flag sounds like |
Walk me through your reasoning for choosing this layout for our main screen. What did you try first? | Name alternatives they rejected and why, tied to your users | "It's clean and modern" |
Where did you use AI on this project, and where didn't you? | Specific and relaxed: "I explored layout options with AI, then rebuilt the chosen one in our component library" | Defensive, vague, or "I don't use AI at all" when the file history says otherwise |
What happens on this screen when there's no data, or the request fails? | Shows you the designed state | "We can add that later" |
Who is this screen for, and what's the one thing they need to do here? | Name your actual customer and their job | A generic persona or a list of every feature |
If we change the brand color tomorrow, how many places will you need to edit? | "One. It's a variable." | "A few hours of updates" |
Ask these without accusing anyone. Good designers welcome the conversation; they wish more clients started it. If you're still choosing who to work with, use these questions in your interviews. Our guides on hiring a SaaS designer and choosing a UX design agency cover the rest of the vetting process.
Test 5: The logo swap test
Pull up your homepage next to your two closest competitors. Now swap the logos, or do it in a screenshot. If nothing feels unusual, your design isn't doing any positioning work, no matter how polished it looks.
This matters more than any single visual tell. A buyer comparing four vendors needs to remember you in three days. When everyone generates from the same averages, sameness becomes the default. We've written about why so many SaaS products look the same and how to break out without hurting conversion.
What is not evidence
Before you confront anyone, know which signals are weak.
AI detector scores. Image and text detectors are unreliable on interface design, and none of them can see process. Never base a decision or an accusation on a percentage.
One tell in isolation. Inter, a gradient, or a card grid alone proves nothing. Conventions that reduce friction, such as a familiar pricing table or a standard nav, are good design.
Speed. Fast delivery from a senior designer who uses AI well is exactly what you want.
Polish. Slop is usually polished. That's what makes it hard to spot.
Honesty about AI use. A designer who tells you where they used AI is showing a green flag, not a red one.
Ownership, contracts and disclosure
There's a business reason to care beyond taste. In its January 2025 report on copyrightability, the US Copyright Office said that material generated entirely by AI isn't protected by copyright and that prompts alone generally don't give enough control to make someone the author of the output. Human selection, arrangement, and modification of AI output can be protected. In plain terms: the more of your product UI and brand that is unedited AI output, the thinner your ownership claim over it may be. That can come up in fundraising diligence or an acquisition.
This isn't legal advice, so run your agreement past a US attorney. When you do, these are the points worth raising:
A short written note at delivery on where AI was used and who reviewed it.
IP assignment that covers human-authored work and all source files.
Delivery of editable source (Figma files, tokens, component code), not flattened exports.
A warranty that AI tools were used in accordance with their commercial terms.
Independent trademark screening for any AI-assisted logo or wordmark.
The founder's AI slop scorecard
Give one point for each check that passes. This is how we review products, not with a scientific score, but by turning a vague feeling into something you can discuss.
No. | Check | Passing Looks Like |
1 | Typography | A clear size scale and at least one deliberate type choice |
2 | Color | Primary, danger, and success colors are consistent and predictable |
3 | Hierarchy | The main action on each screen is obvious within two seconds |
4 | Copy | Specific to your product and customer; no placeholders |
5 | Empty states | A new user knows exactly what to do first |
6 | Errors | Messages explain what went wrong and how to fix it |
7 | Consistency | The same action looks and behaves the same everywhere |
8 | Mobile | Usable on a real phone, not just a resized window |
9 | Source files | Components, named styles or variables, real version history |
10 | Rationale | The designer explains three choices and their AI use clearly |
8 to 10: you have a designer who uses AI well. Keep them. 5 to 7: there's slop in places, usually fixable with a focused review. 0 to 4: much of the work is likely unreviewed AI output. Talk to your designer before the next sprint, and get an independent review before building more on top of it.
You found slop. Now what?
Talk before you fire. Share the scorecard. Sometimes a designer under deadline pressure ships first drafts as finals. A clear standard usually fixes that in a sprint.
Triage instead of redesigning. Most AI slop doesn't mean starting over. Sort issues into keep, fix, and rebuild. Our guide to AI UX debt shows you how to decide.
Put a system underneath. Tokens and components stop defaults from creeping back the next time anyone, human or model, generates a screen.
Check the build too. If your product was generated with tools like Lovable, v0, or Cursor, design is only half the risk. Here's how to audit a vibe-coded MVP before you scale.
Get an additional set of eyes when the stakes are high. Before a raise, launch, or big hire, an outside review is cheaper than building on the wrong foundation.
The bottom line
AI slop isn't about whether a machine touched your design. It's about whether someone who understands your product made the decisions. Run the five tests, fill in the scorecard, and have one honest conversation with your designer. You'll know more in 30 minutes than any detector can tell you.
If you want someone experienced to do this with you, Foundey's 5-day product design audit runs every flow through five lenses: Mental Effort, Conversion Friction, Trust Architecture, Information Hierarchy, and Feedback Channels, and gives you a prioritized roadmap for fixes. If you need design judgment inside your team every week, check out how our embedded product design team works.
FAQ
Is it bad if my designer uses AI?
No. Used well, AI makes good designers faster at exploration and production. The risk is unreviewed output: screens shipped without anyone deciding whether they fit your users, your brand, and your edge cases.
Can an AI detector tell me if a design was AI-generated?
Not securely. Detectors are built mostly for images and text; they struggle with interface design and can't tell whether a human reviewed the work. The tests in this checklist tell you more.
What should I ask a designer about their use of AI?
Ask where they used AI and where they didn't, why they chose the layout of your main screen, and what happens in empty and error states. Clear, specific answers are a good sign.
Who owns a design made with AI?
Under current US Copyright Office guidance, purely AI-generated material isn't copyrightable, while human-authored selection and modification can be. Ownership between you and your designer depends on your contract, so have an attorney review your IP assignment.
Can AI slop be fixed without a full redesign?
Usually, yes. Most fixes target typography, color meaning, empty and error states, and consistency, built on a small design system. A full redesign is only warranted when the information architecture itself is wrong.
How long does a proper design review take?
The self-check here takes about 30 minutes. A thorough expert audit of a SaaS product typically runs about a week. Foundey's audit is a 5-day sprint.
Disclosure: Foundey uses AI tools for research and production every day. Every decision that ships is made and reviewed by a senior designer.
