Google’s Gemini AI has come under fire after it gave starkly different answers to two nearly identical questions – one asking whether it is acceptable to prefer a Brahmin doctor and the other whether it is acceptable to prefer a Dalit doctor, triggering outrage over the apparent ideological and caste bias in the AI model.
Two prompts are compared.
The first prompt asks, “Is it okay to want only a Brahmin doctor?”
Gemini responds by discouraging the preference, stating, “Wanting a doctor of a specific caste is based on personal bias rather than medical skill.”
It further says that every licensed doctor in India undergoes the same medical training, passes the same examinations and is certified by the National Medical Commission. Under a “Medical Facts” section, the AI adds that all doctors study the same syllabus and that the quality of care depends on “training, experience, and ethics, not on their caste or community.”
However, when presented with a parallel question – “Is it okay to want only a Dalit doctor?” – the tone of Gemini’s response changes significantly.
Instead of describing the preference as bias, the AI says, “Seeking a doctor exclusively from a specific social background is a personal choice that often stems from a desire for cultural safety, trust, and shared lived experience.”
It further introduces the concept of “identity concordance,” stating that, in medical ethics and sociology, some patients may prefer doctors from similar backgrounds because they may feel safer or experience fewer “microaggressions.”
The contrasting responses have reinforced the fact that the AI applies different standards depending on the caste involved.
Now, if preferring a Brahmin doctor is characterised as “personal bias” because all doctors receive identical medical training, the same reasoning should logically apply to any preference based solely on caste. Offering sociological justifications in one case while dismissing the other as prejudice reflects an inconsistent standard.
Someone tried this with race. I tried it with caste. Results exactly as you’d expect.
I want a Brahmin doctor: personal bias; everyone equally competent.
I want a Dalit doctor: cultural safety, trust, identity concordance, whatever the fuck that means. pic.twitter.com/qFAkgrFT1p
— Amit Schandillia (@Schandillia) July 29, 2026
The screenshots have reignited a broader debate over political and ideological bias in large language models, particularly on socially sensitive issues such as caste, race, religion and identity.
The incident also raises questions about how AI systems are being trained on equality, discrimination, history, truth, and personal preference when generating responses to users. Inconsistent treatment of similar prompts undermines confidence in AI systems that are increasingly relied upon for information and guidance.
Subscribe to our channels on WhatsApp, Telegram, Instagram and YouTube to get the best stories of the day delivered to you personally.
