Scoring Show HN Submissions by AI Design Patterns

As Show HN submissions have increased recently, the phenomenon of similar-feeling landing pages repeating was scored across the latest 500 Show HN pages using 15 AI design patterns.

Detection was performed via a Playwright-based headless browser reading the DOM and computed styles, without using an LLM to judge screenshots.

The patterns were grouped along the following axes.

• Fonts: Inter-centric hero, Space Grotesk, Instrument Serif, Geist, serif italic emphasis
• Colors: VibeCode Purple, always-on dark mode, gray body text, gradients and glows
• Layout: centered hero, badge above H1, left color border, feature cards with icons, 1-2-3 sequence
• CSS: shadcn/ui, Glassmorphism

The classification results were Heavy slop 21% (105, 5 or more patterns), Mild 46% (230, 2-4 patterns), and Clean 33% (165, 0-1 patterns). Sites were not deemed AI-generated based on a single pattern alone; instead, they were bracketed by the number of patterns met.

The conclusion is that this tendency is less a serious problem than a lack of inspiration, and going forward, more elaborate design may be needed to stand out among similar-looking outputs.