This chapter examines the most common sources of failure in human–AI communication and explains why misunderstandings arise even when users act in good faith. Rather than treating these failures as isolated technical defects, the chapter frames them as predictable outcomes of misaligned communication practices, inaccurate mental models and the interpretive nature of generative AI systems. It analyses frequent user errors, model misinterpretations and hallucinations, the effects of ambiguity and implicit assumptions and the risks associated with overtrust and undertrust in AI responses. The chapter also addresses the communicative implications of guardrails, refusals and safety boundaries, showing how these mechanisms shape interaction and user perception. By identifying recurring pitfalls and explaining their underlying causes, the chapter equips readers with the diagnostic awareness needed to recognise, interpret and correct communication breakdowns. This understanding provides a critical foundation for more advanced forms of collaborative and co-creative communication with AI, explored in subsequent chapters.