Where the AI Look Comes From
Purple gradient, frosted glass, rounded cards: AI defaults to the average of its training data, and average is mediocre. Tap the AI-look tells on a typical AI-generated page; collect them all to unlock why
THE QUESTION THIS PAGE ANSWERS
ANSWER FIRSTWhere the AI Look Comes From?
Purple gradient, frosted glass, rounded cards: AI defaults to the average of its training data, and average is mediocre. Tap the AI-look tells on a typical AI-generated page; collect them all to unlock why
Turn taste into a behavior the product can repeat. The useful outcome is not a nice opinion. It is a visible rule, a small example, and a way to tell when the experience falls below the bar.
Capture one before-and-after example that shows the quality bar without extra explanation.
Polish that improves the surface while leaving the user's uncertainty untouched.
Below is a typical AI-generated landing page. It hides five high-frequency AI-look tells—tap them one by one. Each hit explains why that tell became AI's default move.
Trusted by 10,000+ teams worldwide
The AI look had an older academic name. In About Face 4 Chapter 17, Cooper calls this stuff visual noise: extra visual elements that yank attention off what actually carries information. He lists seven forms. This "team weekly admin" hits all seven—check them off against the list.
Among the seven noises, "too many colors" has a proper name. About Face 4: colors crowded like a palette overwhelm users—that's the carnival effect. Max out saturation and the noise doubles, stealing the scene from content. Same weekend-market flyer, two palettes—tap the one people can actually read.
When AI generates design, it picks the high-probability region of training data—the "average" of web design. Same mechanism as language hallucination: in conversation, fluent beats true; in design, common beats good. It ships the purple-gradient trio the way it invents a nonexistent book title with a straight face: both pick the answer that "most looks like it belongs here."
Good news: the mechanism leaves a door open—the probability distribution shifts with input. The more specific your description, the narrower the model's options. Narrow enough, and you pull it off the average.
Same brief, three levels of specificity—watch the output change.
The AI look has a source: models default to high-probability training regions—the average of web design. The purple-gradient trio is that region's storefront.
Visual noise has a pathology list: over-decoration, info-free 3D, heavy separators, crowded elements, dense color/texture contrast, too many colors, weak hierarchy—seven from About Face 4. Colors jammed like a palette even have a name: the carnival effect.
Same mechanism as hallucination: in chat, fluent beats true; in design, common beats good. Wherever you gave no instruction, it fills in the most common answer.
The fix is cranking specificity: a reference plus checkable hard constraints (one primary color, no gradients). Each notch pulls output farther from the default look. Next time before AI generates a page, put these five tells on a ban list.
Source: Original to Xiaoshan Academy's Taste Engineering series; some design principles adapted from About Face 4, Chapter 17.
Turn the feeling in “Try it · Catch the "AI look" first” into a judgment
“Below is a typical AI-generated landing page.” points out that AI has lowered the bar for making something usable. The skill readers need is noticing what is wrong and turning that feeling into an actionable requirement.
Watch the user's next action, not just the surface
Turn “The AI look had an older academic name.” into observable questions: does the user know what happened, what to do next, and how to recover from an empty or failed state? Does the hierarchy make the important information visible first?
Pretty is not the same as usable
Apply “The fix is cranking specificity: a reference plus checkable hard constraints (one primary color, no gradients).” to a second screen or flow. Record one moment of hesitation and the user action after the change; observable behavior is stronger evidence than polish alone.
From “Try it · Catch the "AI look" first” to “Try it · Visual-noise checklist”
“Try it · Catch the "AI look" first” grounds the problem in “Below is a typical AI-generated landing page. It hides five high-frequency AI-look tells —tap them one by one. Each hit explains why that tell became AI's default move”. “Try it · Visual-noise checklist” then moves it toward “The AI look had an older academic name. In About Face 4 Chapter 17, Cooper calls this stuff visual noise : extra visual elements that yank attention off what actually carries information. He lists seven forms…”. Together, they show that the lesson is not just a conclusion to remember, but a claim with conditions.
Carry the judgment into the next situation
For experience work, turn abstract impressions into user actions: did the person understand the state, find the next step, recover from an error, and want to continue?
- “Try it · Catch the "AI look" first”: Below is a typical AI-generated landing page. It hides five high-frequency AI-look tells —tap them one by one. Each hit explains why that tell became AI's default move
- “Try it · Visual-noise checklist”: The AI look had an older academic name. In About Face 4 Chapter 17, Cooper calls this stuff visual noise : extra visual elements that yank attention off what actually carries information. He lists seven forms…
- “The closing point”: Visual noise has a pathology list: over-decoration, info-free 3D, heavy separators, crowded elements, dense color/texture contrast, too many colors, weak hierarchy—seven from About Face 4 . Colors jammed like a…
The final “The closing point” brings the discussion to “Visual noise has a pathology list: over-decoration, info-free 3D, heavy separators, crowded elements, dense color/texture contrast, too many colors, weak hierarchy—seven from About Face 4 . Colors jammed like a…”. The useful thing to carry forward is knowing which judgments must be revisited when input, scale, or risk changes.
I turned one judgment from this article into a small experiment I could run today. Knowing what to observe next is more useful than simply remembering the conclusion.
After reading this, I first looked for the conditions behind the idea instead of copying the method into a project. That order made the later trade-offs much clearer.
When this judgment reaches real work, which constraint should be added first? I am curious which step matters most between reading and the first practical attempt.
No discussion on this article yet.