How reliable are chatbot answers – and how do I deal with hallucinations?

Chatbot responses are often helpful, but not always reliable. Models can invent or misrepresent content because they are based on probabilities rather than verified sources. Additionally, there is the risk of 
prompt injections : manipulated input that inserts false or unwanted content into responses. For companies, this means: regularly verify facts, maintain consistent content, and document any inaccuracies.