Testing 41 shades of blue on AI agents: they really do prefer some colors over others
danshipper · x · 2026-10-21
An experiment built 41 copies of a product page differing only in the Add-to-cart button color, showed each to fresh agent sessions, and asked "Would you buy this?" 50 times per shade per agent.
Findings:
- Claude Sonnet reacted most strongly to color: 3 shades significantly above average, 4 below, with split-half buy-rate correlation r=0.65 (p<0.001) — its preferences are stable and reproducible.
- Claude Haiku: 4 above / 5 below, r=0.61; Claude Opus: weaker, r=0.41.
- GPT-6.1-Sol barely cared: 0 above / 2 below, r=0.26 (p=0.06, not significant).
Takeaway: agents exhibit statistically significant, stable color preferences that differ across models — directly relevant to designing pages meant to convert AI-agent shoppers.
Related event: Experiments Show AI Agents Have Consistent Color Preferences(2 posts)→
More from Fun
- Overheard in Silicon Valley: 'We're profitable if you don't factor in the token spend' — tech__unicorn · 2026-10-21
- Boardy, the AI Networking Bot, Is Launching a Limited-Edition Plush — chickefitz · 2026-10-21
- Jane Street quants could kill your AI central-planning god overnight, argues quant — fkasummer · 2026-10-21
- POV: Everyone's Too Busy vibe-coding Their Own Game to Play Yours — yacineMTB · 2026-10-21
- Grokbot agent drafts industrial control schematics in open-source PanelCAD — burhop · 2026-10-21
- A Philosophical Dictionary for Beginners: Every Fancy Adverb Means 'Very' — burny_tech · 2026-10-21