Using AI as a Writing Feedback Tool: Prompts That Improve Your Drafts

AI tools have created a new kind of writing resource: a patient, available-at-any-hour reader that can give immediate feedback on clarity, structure, and argument. Used well, AI feedback significantly reduces the gap between what you intended to write and what you actually produced. Used poorly, it either produces generic praise or creates content that is not your own.

The research basis for AI-assisted feedback

Kasneci et al. (2023) in Learning and Individual Differences identified feedback as one of the highest-value applications of AI in education — because it is a use case where AI supplements rather than replaces student cognitive work. The student writes; the AI responds with observations; the student decides how to act on them. This preserves the learning value of the writing process while providing the kind of formative feedback that is often unavailable outside of tutor office hours.

Yan et al. (2024) in the British Journal of Educational Technology distinguish between AI use that offloads thinking (problematic for learning) and AI use that scaffolds thinking (educationally valuable). AI feedback falls in the scaffolding category when the student retains the intellectual work of deciding what to change and how to change it.

What AI feedback can and cannot do

AI feedback is reliable for:

• Identifying unclear sentences and passages
• Flagging structural problems (paragraphs that don't have a clear claim, conclusions that summarise rather than synthesise)
• Identifying gaps in argument (sub-claims that aren't supported, counterarguments not addressed)
• Detecting over-long or confusing sentences
• Checking whether your thesis is specific and arguable

AI feedback is unreliable for:

• Evaluating the accuracy of your disciplinary knowledge (AI does not know your subject matter with certainty)
• Assessing the quality of your sources (AI cannot verify what you say about specific papers)
• Evaluating the correctness of your referencing (AI will often confirm incorrect formats if they are plausible)
• Determining whether your argument is conventionally acceptable in your discipline (disciplinary norms vary significantly)

Always verify disciplinary claims and referencing independently — AI feedback is a starting point, not a final verdict.

Effective feedback prompts by type

Structural feedback

For thesis evaluation:

"Here is my thesis statement: [paste thesis]. Is it specific and arguable, or does it state something obvious that most people would agree with? What is the strongest objection someone could raise to this position?"

For argument flow:

"Here are my body paragraph opening sentences in order: [list them]. Does this sequence of sub-claims follow a logical progression? Are there gaps in the argument or sub-claims that don't obviously connect to the thesis: [paste thesis]?"

For reverse outline:

"For each of these paragraphs, write one sentence summarising what it actually argues (not what the topic is, but what it claims): [paste body paragraphs]."

Analytical feedback

For Explain step quality:

"I'm going to paste a body paragraph. Please tell me whether the 'Explain' step — the interpretation of what the evidence proves — is doing real analytical work, or whether it is just restating what the evidence says. [paste paragraph]."

For counterargument handling:

"My thesis is: [paste thesis]. What is the strongest counterargument someone in [field] would make against this position? Have I addressed it adequately in this passage: [paste relevant section]?"

For evidence-claim connection:

"Does the evidence in this paragraph clearly support the claim made in the Point sentence? Or is there a logical gap between the evidence and the claim? [paste paragraph]."

Clarity feedback

For sentence-level clarity:

"Read this paragraph and identify any sentences that would be unclear to a reader unfamiliar with the topic. For each unclear sentence, explain what specifically is unclear — but do NOT rewrite it for me. I want to understand the problem so I can fix it myself. [paste paragraph]."

For academic register:

"Are there any phrases in this paragraph that sound too informal or colloquial for academic writing? [paste paragraph]."

Final review prompts

Comprehensive structural review:

"I'm going to paste my complete essay introduction and body paragraphs. Please evaluate: (1) Does the thesis make a specific arguable claim? (2) Does each body paragraph open with a clear sub-claim? (3) Is the sequence of sub-claims logical? (4) Are there obvious gaps in the argument or counterarguments not addressed? Do not give general praise — I want specific issues identified. [paste essay]."

How to act on AI feedback

AI feedback identifies problems; you provide the solutions. This is the most important principle.

When AI flags an unclear sentence: ask yourself why it is unclear, then rewrite it in your own words with that understanding. Do not paste the AI's suggested rewrite — you learn nothing from it, and the replacement sentence may introduce academic register problems or lose specific analytical content.

When AI identifies a missing counterargument: research the counterargument in your sources. Do not have the AI generate the counterargument paragraph — that is the intellectual work that academic writing is designed to develop.

When AI says your argument has a logical gap: work through the gap yourself. The experience of identifying why the argument is incomplete and filling the gap is precisely what builds analytical thinking.

The Academic Writing Fundamentals course covers the full essay-writing system including argument planning, PEEL paragraphs, and editing — which gives you the framework to interpret and act on AI feedback effectively. The Essay Structure Planner helps you build your argument before writing, reducing the structural problems that AI feedback is most useful for catching.