AI for Writing Critique: Get Feedback You Can Use

Ask an AI to critique your writing and you will usually get a mix of praise and vague suggestions. That is the default: be encouraging, point out some improvements, avoid saying anything too harsh. To get feedback you can act on, you need to ask for something specific.

Tools like fenbs let you save the critique prompts that produce useful feedback and the lessons from using them, so your next writing session starts with the version that works rather than a retyped attempt from memory.

Why AI critique is often too polite

AI writing assistants are trained to be helpful and avoid responses that feel harsh. Applied to critique, this tendency produces feedback that identifies minor issues without naming the main structural problem, or gives suggestions framed as options ("you might consider...") rather than findings ("the argument breaks here because...").

The fix is not to ask the AI to be harsh: it is to ask for evidence. "What specific sentence in this draft undermines the main argument?" requires a finding. "What do you think of this?" invites reassurance.

How to ask for structural feedback

Structural feedback identifies problems with the argument, the order of ideas, and whether the draft does what it claims to do. Ask for it separately from line-level feedback.

• "What is the main argument of this draft? State it in one sentence."
• "Which section does the least work to support that argument?"
• "Where does the draft assume the reader already knows something it has not explained?"
• "Is the conclusion supported by the evidence in the draft, or does it introduce new claims?"

Each of these is answerable from the text. The AI cannot reassure you about structure the way it can reassure you about quality: it has to engage with what is actually there.

Run structural questions at two stages: on the outline before writing, and on the finished draft. Structural problems visible at the outline stage are easier to fix than after the prose has accumulated. A draft that changes direction at the halfway point is more expensive to revise than an outline that does the same. AI feedback on an outline takes seconds and can redirect the work before it is too costly to restructure.

Prompts for style and sentence-level critique

Style feedback is most useful when it targets specific problems rather than general tone. Name the issue you want checked:

• "Find every sentence over twenty-five words and rewrite it as two shorter ones."
• "Find every instance of passive voice and identify which ones weaken the sentence."
• "List every hedge phrase ('might', 'perhaps', 'it could be argued') and flag the ones I can remove."
• "Find the three weakest transitions between paragraphs and suggest a replacement for each."

These prompts produce a list you can work through. "Improve the writing style" does not.

Checking the AI's evidence

Critique feedback is only useful if it is based on the text, not on the AI's assumptions about what you were trying to write. For any piece of feedback that names a specific problem, check that it cites the relevant sentence or section.

Add "quote the relevant sentence" to your critique prompt. If the AI cannot quote a sentence that supports its finding, the finding may be an inference about the reader's likely reaction rather than a claim about the text. Both are useful, but they are different kinds of feedback.

Save critique prompts that produce reliable, evidence-based feedback in fenbs so the same prompt is available for the next draft or the next writing session. Link the version to the lesson about what the earlier version got wrong.

How to do this with fenbs

1. Create a free account and add a project for the writing work.
2. Add the critique prompts to Skills and prompts in fenbs: one for structure, one for style, one for the evidence check.
3. After each critique session, record lessons in Lessons learned: which prompt produced the most useful feedback and what you changed as a result.
4. Connect your AI app over MCP from Integrations so it reads the current critique prompts at the start of each session.

A note on AI critique and your own judgment

AI critique is a starting point, not a verdict. The AI does not know your audience's specific expectations, the context that makes a paragraph necessary, or the strategic reason an argument is phrased a certain way. Use critique feedback to surface structural and stylistic candidates for review, then decide for yourself.

One useful sequence: run structural prompts first, make any changes the findings justify, then run style prompts on the revised version. Mixing structural and style feedback in a single pass produces a longer list that is harder to prioritize. Structural changes often eliminate the style problems they were covering: fix the argument before polishing the wording, and the polishing step is often much shorter as a result.

For AI writing tools more broadly, how to use ai at work covers the question of which tasks AI handles well and where review matters most.