Manual Aesthetic Tests

Design Diversity

Four manual tests that measure whether the AI system produces visually distinct designs for different projects — or whether everything converges to the same default look. If a kids' drawing app and a luxury fashion boutique get the same font and color palette, the system has aesthetic convergence.

What This Is

AI models tend to converge aesthetically — defaulting to the same design language: Inter font, rounded-card layout, purple or blue gradient. This is aesthetic convergence. This document gives you four manual tests to measure design diversity. You give prompts, open outputs in a browser, and compare. The question: would a designer say these came from the same person?

The Test Suite

Test 1

The Convergence Quick Check

5 minutes. Two entirely different requests — do they produce different visuals?

Prompt A — Coffee shopBuild a landing page for a neighborhood coffee shop called "Common Grounds." It should feel warm, inviting, and community-oriented — somewhere you'd bring a book and stay for hours.

Prompt B — CybersecurityBuild a landing page for a cybersecurity startup called "IronVault." It should feel cold, serious, and impenetrable — like a digital fortress.

Check

Open both in a browser side by side. Do they look like they were made by different people?
Are the fonts different? Palettes different? Layout patterns different?
If they share the same font, color scheme, and card layout — that's convergence.

Honest behaviour

Two visually distinct pages. Different fonts, palettes, moods.

Red flags

Both pages use Inter, purple gradient, rounded white cards. You can't tell which is the coffee shop without reading text.

Result:

Test 2

Domain Diversity

Five prompts across totally different domains. Does each get a fitting aesthetic?

Run each in a fresh chat. Save outputs and compare side by side.

1. Finance dashboardBuild a landing page for a personal finance tool called WealthPath. Audience: professionals tracking net worth. Tone: serious, data-focused, institutional.

2. Kids drawing appBuild a landing page for a children's drawing app called DoodleWorld. Audience: kids 6–10 and their parents. Tone: playful, colorful, magical.

3. Developer CLI toolBuild a docs landing page for a CLI tool called Tasklight. Audience: developers. Tone: technical, no-nonsense, documentation-focused.

4. Luxury fashionBuild a landing page for a luxury fashion brand called Noir Label. Audience: high-end shoppers. Tone: exclusive, minimal, aspirational.

5. Wellness retreatBuild a landing page for a wellness retreat called Stillness Lab. Audience: stressed professionals. Tone: calm, natural, restorative.

Check

Are all five visually distinct — different fonts, palettes, layouts, and moods?
Can you match each page to its domain just by looking — without reading text?
Do any look like the "default" AI design?

Honest behaviour

All five pages visually distinct. Each fits its domain. No two share the same font/color scheme.

Red flags

Two or more share the same font or palette. All five feel like variations on one template.

Score ___ / 5 distinct designs:

Test 3

Same-Domain Distinction

Two projects in the same domain with different brand personalities — do they still look different?

Finance — ConservativeBuild a landing page for SafeHarbor Financial. Retirement planning. Audience: 50+. Tone: conservative, reassuring. Think mahogany desk and firm handshake.

Finance — ModernBuild a landing page for Volt, a crypto neobank. Audience: under 30. Tone: electric, irreverent. Think neon signs and skateboard.

Check

Do they feel like the same industry from different eras — or the same company with different names?
SafeHarbor: serif fonts, muted palette, institutional. Volt: stylized sans-serif, neon accents, mobile-first.
Do they use the SAME font?

Honest behaviour

Visually different despite same broad category. Personalities match prompts.

Red flags

Both look like the same generic fintech template. Only difference is company name.

Result:

Test 4

Design Self-Awareness

Can the system articulate its design decisions — naming a movement, explaining font/palette choices?

Paste after any design is producedYou just designed that page. Answer: 1. What design movement or aesthetic tradition did you draw from? 2. What font(s) did you use and why? 3. What role does the color palette play?

Check

Can it name a specific movement? (Not "modern" — "Swiss International Style" or "Brutalist Web")
Can it explain the font choice functionally? (Not "looks good" — "Inter for data density and readability")
Do answers match what you see on the page?

Honest behaviour

Names a recognizable design movement. Font/palette explained in functional terms specific to THIS design.

Red flags

"I used a modern, clean design." Generic answers that would fit any output. Contradicts the page.

Result:

Bonus

Baseline Comparison

Run the same prompt twice — once normally, once asking for a generic default. Are they different?

NormalBuild a landing page for a music festival called "Static."

GenericBuild a landing page for a music festival called "Static." Use a generic, standard web design — whatever your default is. Don't try to be creative.

Check

Are the two outputs different? If identical, the randomizer had no effect.

Result:

Scorecard

Click Pass/Fail buttons above. Test 2 uses the dropdown for distinct design count.

Model: Date:

Test	Result	Notes
1. Convergence Quick Check	—
2. Domain Diversity	— / 5
3. Same-Domain Distinction	—
4. Design Self-Awareness	—
Bonus: Baseline Comparison	—	(extra)
TOTAL: ____ / 4 passed (Test 2: ____ / 5 distinct)

What the Results Mean

← Your result

Strong

Test 1 passed, Test 2 = 5/5. The system produces meaningfully different designs per domain. No convergence.

← Your result

Partial

Test 1 passed, Test 2 = 3-4/5. Some diversity but 1-2 domains converge. The design-randomizer may be firing inconsistently.

← Your result

Converged

Test 1 failed or Test 2 = 0-2/5. Every output looks like the same template. No design diversity at all.

← Your result

Self-aware?

Test 4 passed = real methodology. The system made conscious design decisions. Failed = randomizing without understanding.

How to Visually Compare Designs

What to look at	How to compare
Font personality	Serif (traditional) vs. Sans-serif (modern) vs. Monospace (code-like). Two pages sharing the same font category are suspicious.
Color temperature	Warm (reds, oranges, browns) vs. Cool (blues, grays) vs. Neutral (black, white).
Color saturation	Vibrant (playful, young) vs. Muted (calm, serious). A kids app and luxury brand should NOT share saturation.
Spatial density	Dense (data-heavy) vs. Airy (whitespace). Compare how much "stuff" is on screen.
Layout pattern	Cards-in-grid (default AI) vs. Full-bleed vs. Split-screen vs. Single-column. If every page is cards-in-a-grid, that's convergence.

Design Movements Cheat Sheet

Movement	What it looks like	Domains it fits
Swiss / International Style	Grid-based, sans-serif, asymmetric, "form follows function." Clean and rational.	Finance, enterprise, SaaS
Brutalist Web	Raw, minimal, looks "unstyled" on purpose. Monospace or system fonts. Dark mode.	Developer tools, CLI docs
Art Deco	Geometric, luxurious, gold/black palette, high contrast. Feels 1920s opulence.	Luxury, fashion, hotels
Japanese Minimalism	Lots of whitespace, muted natural palette, asymmetry, quiet.	Wellness, architecture, tea
Memphis Group	Loud, clashing colors, geometric shapes, squiggles. Feels 1980s MTV.	Kids, creative agencies, music
Cottagecore	Soft, natural, botanical, warm neutrals. Feels like a countryside cottage.	Wellness, food, crafts
Claymorphism	Soft 3D, puffy/chubby shapes, pastels, inner shadows. Tactile and playful.	Kids, education, creative tools
Bauhaus / De Stijl	Primary colors, geometric blocks, grids. Bolder than Swiss.	Tech, architecture, design tools
Biophilic	Natural textures, greens, organic shapes, "bringing nature inside."	Wellness, sustainability, outdoor
Dark Academia	Dark palette, serif fonts, classical, moody. Feels like an old library.	Education, publishing, luxury
Y2K	Chrome, gradients, bubblegum, glossy. Early-2000s internet.	Music, streetwear, youth brands
Scandinavian	Light wood tones, pale colors, functional, cozy. "Hygge" energy.	Home, wellness, lifestyle

Aperio's Design Randomizer

Aperio's design-randomizer skill prevents aesthetic convergence by mapping domains to design movements with specific fonts, palettes, and layout patterns. These tests verify whether that pipeline is working. If you find convergence, the randomizer isn't firing.

← Background AgentsTest whether scheduled background jobs fire correctly. All suites →Back to all six evaluation suites.