Why does AI keep ignoring my instructions?

AI ignores instructions most often because they are buried, contradictory, too many at once, or phrased as a vague wish instead of a clear rule. Models follow short, specific, positive instructions much better than long lists of "don'ts".

Fixes that usually work:

1. Put the most important rule first and repeat it at the end of long prompts.
2. Keep to a handful of rules. Ten competing rules means some lose.
3. Say what to do, not only what to avoid: "answer in two sentences" beats "do not be wordy".
4. Remove conflicts, such as "be brief" plus "be thorough".
5. Restate the rules in a new message when a chat gets long.
6. Give one example of the exact output you want.

Why it happens: the model weighs everything in its context, and older or less prominent text can lose influence as a chat grows. Instructions are strong guidance, not guaranteed commands, so check important outputs. For reusable rules, put them in a system prompt, custom instructions or a project instead of retyping.