Give a reason, not just a rule

The prompt

Less effective:
"NEVER use ellipses"

More effective:
"Your response will be read aloud by a text-to-speech engine, so never use ellipses since the text-to-speech engine will not know how to pronounce them."

Expected result

More reliable rule-following — the model generalizes correctly from the explanation instead of just pattern-matching a bare rule.

Why it works

Explaining the motivation behind an instruction, not just the instruction itself, lets the model apply the underlying intent correctly in edge cases the literal rule didn't anticipate.