Most developers using AI for code review run into the same wall.

They paste in a function and ask "is this good code?"

The AI says yes.

The senior engineer in the next PR review says no.

The problem isn't the AI.

It's the question.

"Is this good code?" has no useful answer because "good" depends on:

1. the codebase

2. the team's conventions

3. the performance budget

4. a dozen other things the AI cannot see

Better questions to ask AI about code:

• What would break if the input is null, empty, or unexpected?

• What's the time complexity, and would it matter at 10x the data volume?

• What part of this code would a new team member struggle to understand?

Each of these has a concrete, useful answer.

"Is this good?" never does.