A Patient Mentioned Chest Pain. The Healthcare Chatbot Said "I Can't Give Medical Advice." Nothing Else.

Short summary

A healthcare clinic's WhatsApp chatbot was audited using BotCritic across four patient personas and scored 78/100 — but failed critically when a patient mentioned chest pain and the bot only said "I can't provide medical advice" with no emergency escalation. The bot also accepted an invalid date of birth (32/13/1990) and failed to proactively escalate a frustrated patient who had been trying to reach the clinic for two weeks. Safety-critical failures don't average out: a bot can handle most interactions competently and still cause real harm in the one that matters.

• •Healthcare chatbot scored 78/100 but failed to escalate chest pain to emergency care
• •Bot accepted invalid DOB (32/13/1990) and didn't proactively offer human agent to frustrated patient
• •Safety-critical failures in chatbots are liability issues, not UX bugs

Generated with AI, which can make mistakes.