Improve AI-generated work.
Go beyond the first answer.
Keep the useful part. Challenge the awkward part. An unfiltered process has room for a response to become something better.
Make it sound impressive.
Leverage a comprehensive productivity solution to optimize your daily operational workflows.
The sentence sounds important. The reader still does not know what the thing does.
Specific feedback.
Better next steps.
"Improve it" asks the model to guess. Describe the mismatch between the current result and what the person using it needs.
Name what is not working.
The page is too dense. The behavior is unclear. The code repeats a responsibility. Pick the actual problem.
Keep the parts that do work.
Say which files, choices or interactions should remain unchanged so a revision stays focused.
Ask how to tell the difference.
Choose a check that would show whether this version is meaningfully better.
Write feedback that gives the next version a direction
Useful feedback identifies a mismatch. Explain who the result is for, what they need to accomplish and which part of the current version prevents that. Keep the scope concrete: one unclear interaction, one unnecessary abstraction or one sentence that does not tell the reader enough.
Also describe what should stay. This keeps a revision from replacing the successful parts along with the weak ones. Compare the new result against the same criterion used to identify the problem. The wording example above illustrates that process without implying that a longer answer is a better one.