7 Easy-to-Avoid AI Website Mistakes
▼ Summary
– Y Combinator’s Aaron Epstein and Cron founder Raphael Schaad identified seven common mistakes in AI-designed websites, including generic design trends like purple gradients and bento-box layouts.
– Unexpected user interaction feedback, such as cursor-following lights and superfluous animations, breaks user expectations and creates a poor experience.
– Broken UX patterns like scroll hijacking and non-standard navigation cause confusion and friction, distracting from the product’s message.
– Weak messaging and poor information hierarchy fail to clearly communicate the product’s value, with AI often adding decorative but non-functional elements.
– The core issue is over-reliance on AI without experienced judgment, leading to generic, inconsistent designs that lack brand identity and strategic direction.
Y Combinator general partner Aaron Epstein recently sat down with Raphael Schaad, the founder of Cron (later acquired by Notion), to break down the most frequent errors they see in AI-generated websites. Their conversation identified seven recurring mistakes that “vibe coders” make, and understanding them can save any startup from launching a site that looks polished but fails to perform.
The discussion began on a positive note. Epstein and Schaad agreed that the ability to vibe code a website is genuinely empowering, and a non-designer can produce solid results. But as they clicked through example after example, a clear pattern of pitfalls emerged.
1. Generic Design Trends
The most common mistake is letting the AI dictate the entire aesthetic. When a user prompts an LLM to “make it look modern,” the output is predictably generic. The design might appear sleek in isolation, but it lacks any brand identity. It feels familiar, unoriginal, and forgettable.
Schaad pointed to the overuse of purple gradients and bento-box grid layouts as hallmarks of this trend. Epstein highlighted the classic AI-generated software dashboard, complete with red, green, blue, and purple callout boxes. “Every fake dashboard looks basically like something like that,” Schaad remarked. The problem, Epstein explained, is not that these design choices are bad on their own. They become a liability when every other startup uses them, stripping away any sense of originality.
2. Unexpected User Interaction Feedback
Good interaction design provides clear feedback. A button should feel clickable, and a click should produce a visible response. AI-generated sites often break this contract with unexpected or confusing interactions.
The panel listed several offenders: lines that follow the user’s cursor, lights that track mouse movement, superfluous background animations, automatic fade-ins, and hover effects that shift elements without purpose. These gimmicks distract rather than guide. They create a disorienting experience, similar to walking through a lobby and bumping into a glass wall that shouldn’t be there.
3. Broken or Confusing UX Patterns
This category covers interactions that make a page harder to use. The biggest culprit was scroll jacking, where the website hijacks the browser’s native scrolling to trigger animations. Epstein and Schaad encountered this four times during their review.
In one instance, a page froze entirely as an animation loaded. “We’re locked into this position,” Schaad noted, questioning what the animation was supposed to communicate. Epstein added that the flashy effects drew all his attention away from the actual product copy. Other issues included non-standard navigation, menus that jump inconsistently, and hover-only interactions that hide key functionality.
4. Weak Messaging and Product Explanation
A stunning design is useless if visitors cannot understand what the product does. The podcast identified sites with unclear value propositions, vague explanations, and too little useful information above the fold. This is often AI slop content that uses ambiguous, lazy language.
A landing page is a customer acquisition channel. If someone cannot quickly grasp the product and its value, the design has failed, no matter how pretty it looks.
5. Poor Information Hierarchy And Structure
AI can generate pages that look structured but are actually cluttered. The panel noted sites with too many competing visual elements: multiple text styles, extra labels that add no meaning, and decorative elements that waste vertical space.
The key insight here is that AI tends to produce content that looks busy but accomplishes nothing. Every word and visual element should serve a purpose. If it doesn’t help the reader process the page, it is noise.
6. Inconsistent Brand and Visual System
A cohesive brand identity is hard to fake. The panel found sites where visual language shifted between sections. Colors did not coordinate with the logo, product screenshots clashed with the landing page style, and different sections looked like they were generated separately.
These inconsistencies signal that the AI was prompted to create something “modern” without any underlying brand strategy. The result feels inherited rather than intentional.
7. Lack of Experience-Based Judgment and Over-Reliance on AI
This final point ties all the others together. AI removes technical barriers, but it does not replace design judgment. The quality of the output depends entirely on the quality of the prompt and the human’s ability to edit.
The panel listed several behaviors that lead to poor results: accepting all AI changes without question, outsourcing taste to the LLM, letting AI decide brand direction, and spending saved time on more effects instead of clearer thinking. The human is now the editor, and that role requires experience and expertise.
The takeaway is clear. AI tools are powerful, but they are not a substitute for strategic thinking. A successful website still requires someone who understands what serves the visitor and the business goal. Without that judgment, even the most technically impressive site will fall flat.
(Source: Search Engine Journal)
