Large-language models increasingly mediate how people seek information, make decisions, and even receive care from social robots. Yet these systems still issue fluent but unfounded answers - "confabulations" that erode trust and, in embodied agents, can pose direct safety risks. We argue that a lightweight, five-step Cognitive-Behavioural Therapy (CBT) loop—inserted inside or immediately above every system prompt—offers a practical defence. The loop forces the model to state its automatic thought, challenge itself, and re-frame with calibrated uncertainty. Recent leaks of Grok's ideology prompt and Anthropic's safety prompt highlight how much behaviour hinges on this hidden layer; our proposal turns that layer into a structured, clinically grounded self-check. Because the loop is model- and platform-agnostic, adds little latency or cost, and grows more critical as model internals become opaque under computational irreducibility, we call on developers to adopt therapy loops as standard practice across chatbots, APIs, and social robots.