Which of these is most likely to improve the quality of an AI assistant's response?

A model cannot ask what you meant before it answers, so everything you leave out it has to guess — and a guess averaged over everything it has read is exactly what a generic answer is. Anthropic's own framing is to treat Claude as a brilliant new colleague who has no context on your norms: capable, but unaware of your audience, your house style or the length you need.

Being specific is what removes the guessing. Name the audience, the format, the length and the constraints that matter. "A 200-word summary for a non-technical manager, as three bullets" asks for something the model can actually hit; "summarise this" does not. Say what you want rather than what you want avoided, because a prohibition still leaves every other option open.

The other three answers all withhold something: the detail, the examples, the format. Each one hands a decision to the model that you were in a better position to make.