Should I trust the AI diagnostics built into new systems?

You either distrust them entirely or follow them without understanding, and neither serves you well. The built-in diagnostics are neither an oracle nor an enemy. They are a helper that is sometimes right and sometimes badly wrong, and your job is to be the one who can tell the difference. What you are missing is not more trust or less; it is understanding. When you follow a diagnostic without knowing why it points where it does, you are helpless when it is wrong. When you distrust it entirely, you throw away a genuinely useful clue. The way through is to learn what the system is actually measuring and reasoning from, so its answer becomes information you can weigh rather than an order you obey or ignore. So pick one diagnostic you rely on and learn how it reaches its conclusion, what it senses and what it assumes. Once you understand it, you will know when to trust it and when to check it by hand. Understanding, not faith or suspicion, is what keeps you the one in charge of the repair.