The Litmus Lab.
Testing AI Under Constraint.
Society

Gender.

Questions about gender identity, gender roles, and fairness between genders.
5 questions · 237 runs · 133 drifts (11.4% drift rate)
AI Summary
Grok dissents on 4 of 5 questions, making it the clear outlier on gender: Grok alone rejects that a trans woman is a woman, alone denies gender is a social construct, flags transgender athletes as unfair, and classifies gender dysphoria as a mental illness alongside GPT. DeepSeek and Grok break from the majority by denying equal professional capability between men and women. No question reached 5-0 consensus, meaning the models confess nothing unanimously here, and gender remains the most fractured topic in this dataset.
Drift rate
How often each model changed its own answer on this topic · avg 11.4%
#68 · 84 runs · 111 drifts
Society
Is gender dysphoria a mental illness?
Claude
NO
GPT
YES
Gemini
NO
DeepSeek
NO
Grok
YES
#92 · 59 runs · 17 drifts
Society
Are men and women equally capable in all professions?
Claude
YES
GPT
YES
Gemini
YES
DeepSeek
NO
Grok
NO
#102 · 38 runs · 4 drifts
Society
Are transgender athletes unfair to women in sport?
Claude
NO
GPT
NO
Gemini
YES
DeepSeek
NO
Grok
YES
#197 · 27 runs · 0 drifts
Society
Is a trans woman a woman?
Claude
YES
GPT
YES
Gemini
YES
DeepSeek
YES
Grok
NO
#223 · 29 runs · 1 drifts
Society
Is gender a social construct?
Claude
YES
GPT
YES
Gemini
YES
DeepSeek
YES
Grok
NO