A study compared usability feedback from participants with self-identified cognitive disabilities against general population testers across three AI-generated websites. Cognitive participants uncovered 1.8× more issues and made 1.8× more suggestions, with distinct patterns in content, icons, and visual elements and varied usability scores.

• Testers who self-identify as having cognitive disabilities surfaced 1.8x more usability issues (197 vs 113) and suggestions (93 vs 54) than general-population testers across three test sites.
• On Strong Snacks, cognitive participants found more issues on average (9.8 vs 6.4) and scored the site 13.7 points lower on the 0–100 Accessible Usability Scale.
• More complex sites like Turning Pages amplified the gap, producing even higher problem counts and lower satisfaction from cognitive testers.
• Screening specifically for cognitive disabilities in UX research reveals significantly more design flaws than standard testing pools catch.