ChatGPT Outperforms Humans at Dating Advice in Five Preregistered Experiments (N=1,722)

ipeirotis · x · 2026-09-10

Osborne & Bailey (Scientific Reports, 2025) ran five preregistered experiments (N=1,722) comparing human and ChatGPT-generated advice on quality, effectiveness, and authenticity. ChatGPT's advice beat the average online participant's even in dating and relationships — the domain where people say they most strongly prefer human advice — while documenting a clear bias against AI-labeled advice.

The upshot: aversion to AI advice reflects subjective bias, not objective quality.

Related event: ChatGPT Beats Humans in Blind Evaluation of Relationship Advice(2 posts)→

Original post →

More from Research

Research channel →