ChatGPT Voice in India

Indian students are heavy, mobile-first ChatGPT users, yet few use voice input. The usual assumption is that they can't find it. I set out to test that, alongside four other explanations.

• DiscoverabilityRuled out. 97% already knew voice existed.
• Social discomfortConfirmed. Average comfort was 3.82 out of 5 alone, and 2.06 in a library or shared space.
• Preference for typingStrongly confirmed. 71% still preferred typing, mostly because it's faster (12 mentions) and gives more control (10).
• PrivacyPartly. Present, but secondary to fit with how they work.
• Accent and Hinglish recognitionPartly. A barrier for a few, not most.

"I use voice only when my hands are busy." "I don't want people around me hearing my queries." "For assignments and coding, I still prefer typing."

Stop treating voice as a general alternative to typing, and anchor it to moments where it clearly wins. Suggest it when hands are likely busy or during a commute, and when someone hesitates on an empty text field. Let people edit the transcript before sending, which addresses the control problem directly.

• Discoverability fixesBigger mic buttons and first-launch tooltips solve a problem that 97% of respondents don't have.

Thirty-four students is enough to rule out the discoverability assumption with confidence. It isn't enough to size the opportunity, and it says nothing about non-students.