About AI-generated responses

Responses are generated, not retrieved

Shravya AI produces answers with a large language model. Unless it tells you it searched the web, it is not looking anything up in a verified reference — it predicts a plausible reply, which is usually useful and occasionally, confidently, wrong.

Where it is least reliable

• Medical. General wellbeing conversation only. Not diagnosis, not treatment, not a reason to delay seeing a doctor.
• Legal. Not legal advice, and no substitute for a lawyer who knows your situation and jurisdiction.
• Financial. Nothing here is investment advice or a recommendation to buy or sell anything.
• Current events. The model has a training cutoff and may be confidently out of date.
• Numbers, names and citations. These are exactly what a language model is most likely to invent. Verify them.

Personas are characters

Doctor Dadi, Peace Pandit, Bug Baba, Zindagi Guru and Buddy are tones of voice, not credentials. A persona sounding certain, warm, or authoritative tells you nothing about whether the answer is correct.

Voice calls

On a call your speech is transcribed before the assistant sees it. Transcription can mishear — accents, background noise, and switching between Hindi and English all make it harder — so the assistant may answer something slightly different from what you actually said. If a reply misses the point, it is worth repeating yourself.

You are responsible for what you do with it

Decisions you take on the basis of a response are yours. Where the stakes are real, please check with someone qualified.