TOEFL Academic Discussion — Band 5 Sample (AI in Writing Classes)

Prompt

Your professor has posted the following discussion question, with two student replies already visible.

Professor Okonkwo:
Most universities have now adopted some kind of policy on generative AI tools in coursework, but first-year writing courses remain a contested case. Some instructors argue that allowing AI assistance lets students focus on higher-order skills like argument and revision. Others argue that the early drafts of beginner writers are exactly the work that AI replaces, and that delegating those drafts robs students of the practice that builds writing ability in the first place.

For this discussion: should first-year university writing courses permit students to use generative AI tools? Justify your position.

Priya:
I'm in favour of allowing them, with disclosure. Outside the classroom, professional writers already use AI to brainstorm and edit. A writing course that pretends those tools don't exist isn't preparing students for actual writing in 2026. Required disclosure of AI use keeps the work honest.

Marcus:
I disagree. Beginner writers learn by struggling with sentences they cannot yet produce — that struggle is the curriculum. If a student lets AI generate the first draft, they have skipped the part that actually develops their writing. A 2024 study at the University of Michigan found that students permitted unrestricted AI use scored lower on revision tasks at the end of the term than the control group.

Instructions: Post a substantive 100+ word reply that engages with at least one of the student replies and takes a clear position. Use accurate vocabulary and sentence variety.

Sample response

Marcus's argument is the more honest one in the early stages of learning to write, but I think Priya is right about where the course should land by the end of the term. The disagreement only looks total because the two of them are talking about different points in the same fifteen-week trajectory.

Marcus is correct that struggling to produce a sentence is part of how the skill is built. The Michigan finding he cites tracks with what writing teachers have observed for years: students who outsource the generative phase do not develop the implicit pattern recognition that lets a more experienced writer feel when a sentence is wrong. Cognitive offloading erodes the very ability the course is meant to grow.

But Priya is right that the course also has a downstream goal — to prepare students for the actual writing they will do after graduation, where AI tools are now part of the workflow. Banning the tools entirely solves the early-stage problem at the cost of the later-stage one.

A defensible course design follows from combining their views. AI use should be prohibited in the first six weeks, when students are building the generative habit, then progressively permitted with disclosure for the remaining nine, with assignments specifically designed around AI-assisted revision rather than AI-generated drafts. That way Marcus's concern is honoured early and Priya's is honoured late.

Rubric breakdown

What works in this response

• Reframes the question. Saying "the disagreement only looks total because they are talking about different points in the same trajectory" is a band-5 move — it shows the writer can see the structure of the debate, not just pick a side.
• Engages with Marcus's evidence (the Michigan study) by accepting it and explaining the mechanism, rather than waving it away. Examiners reward responses that are stronger because they have absorbed the counter-argument.
• "Cognitive offloading" is field-specific vocabulary used in context. It signals topic-area mastery without sounding forced.
• Concludes with a concrete proposal (6 weeks / 9 weeks / disclosure / revision-not-drafting). Concrete proposals are far more memorable than abstract conclusions and demonstrate that the writer has thought about implementation.
• Word count comfortably above the 100-word floor. Top-tier responses on this task typically run 200-280 words; 241 sits in that band.

How to push higher

Band 5.0 is essentially the rubric ceiling for this task type. Pushing into 5.5+ territory would require an additional rhetorical layer: a brief acknowledgement of evidence that complicates the synthesis position itself (for example, that the proposed six-week boundary may be unworkable in courses where students arrive at very different baseline levels). Adding that intellectual humility on top of an already strong argument is the only optimisation worth attempting.

Common mistakes at lower bands on this prompt

Band 3 responses on this prompt typically (1) take a hard yes/no position without engaging either peer reply substantively, (2) treat the prompt as a "should AI be banned" question rather than a "what should writing courses do" question, and (3) drift from first-year writing courses into general statements about AI in education. Band 4 responses engage with one peer fairly but the position usually amounts to agreement with that peer rather than a new contribution. The single biggest move from band 3 to band 4 on this prompt is acknowledging the strongest counter-argument visible in the thread (here, Marcus's Michigan finding) before responding to it.

More TOEFL Writing samples

• Academic Discussion — Band 5 sample (sin tax)
• Academic Discussion — Band 3 sample (with critique)
• Email task — Band 5 sample (grading-scheme change)
• All TOEFL Writing samples