The article explains why relying on AI or Large Language Models (LLMs) for tuning ZFS can lead to misleading and potentially harmful configuration advice. Real-world tests reveal that AI often provides outdated, incorrect, or incomplete recommendations due to its reliance on statistical models and training data. ZFS, a complex file system with numerous adjustable parameters, requires a deep understanding of its parameters and their interactions, which AI cannot consistently provide. The article highlights several examples where AI gave inaccurate advice about ZFS parameters, emphasizing the risks of trusting AI for critical system configurations. Instead, it recommends consulting experienced engineers or upstream contributors for reliable ZFS tuning and support.