How to Give AI Feedback That Improves the Next Draft

Why “make it better” is too vague

An AI assistant cannot reliably infer whether “better” means shorter, friendlier, more accurate, less salesy, more specific, or suitable for a different reader. Vague feedback encourages it to guess. That can replace a good detail while leaving the important problem untouched.

Use the four-part feedback stack

Give the next draft four pieces of direction: keep what is already right, change the exact problem, limit anything that should not be added, and check the facts or constraints before returning the revision. This keeps the feedback practical and makes the change easy to inspect.

For example: “Keep the opening sentence and the deadline. Change the middle because it sounds too formal. Limit the message to 120 words. Check that it does not promise a feature we have not released.”

Show the assistant the reason, not only the edit

Instead of replacing a sentence silently, tell the assistant why it was changed. “This phrase sounds like marketing; write in a direct, helpful tone for existing customers” teaches a reusable preference. It also lets you judge whether the assistant understood the principle, rather than merely copying a line you supplied.

Revise one dimension at a time

Fix the structure first, then accuracy, then tone, then length. Asking for all of these at once makes it hard to see what improved and can create new errors. For a high-value draft, save each version or keep a short change log so you can compare the revision with the original.

Know when prompting is no longer the answer

If a draft lacks facts, source material, product knowledge, policy context, or a real point of view, more rewriting will not solve the problem. Pause, gather the missing material, and then provide it or ask the right person. AI can improve language, but it cannot supply trustworthy knowledge that was never in the task.

How to apply this guide

Use this guide when your AI results feel vague, generic, too long, or hard to review. Better prompting is not about secret phrases. It is about designing a clearer task with boundaries and a review path.

• Write the task in one sentence before opening an AI tool.
• Decide which parts need human review: name the specific problem, protect what is correct, ask for one revision goal.
• Remove private or unnecessary context before prompting.
• Check whether the final output changes a fact, promise, number, date, or decision.

A prompt should reduce ambiguity without adding private details the model does not need. If the prompt becomes a long dump of context, stop and separate the source, task, and checks.

A safer prompt to try

Use this starter prompt when you want help with the idea in this guide but still want the model to show its limits.