TOEFL Academic Discussion: the complete guide (2026 format)
One of the two open Writing tasks is a class discussion board: a professor asks a question, two classmates answer it, and you post your own contribution. This guide explains what the task rewards, what the peer posts are actually for, and how to spend the ten minutes.
What the task asks you to do
You are shown an online discussion board from a university course. A professor states a question — usually a policy or preference question with two defensible sides — and two students have already replied, taking different positions. Your job is to add your own post: state where you stand, explain why, and do it in a way that belongs in that conversation rather than beside it.
The setting is not decoration. The task is called Write for an Academic Discussion because the discussion is part of the prompt: the two student posts are material you are expected to use. A response that could have been written before reading them is a response that has ignored half of what it was given.
What ETS states
The current Writing section lists 12 items: 10 Build a Sentence items, one Write an Email task, and one Academic Discussion task.
The 2026 specification lists the Writing section at 12 items and 20 maximum raw points.
The specification states that blueprint values are subject to minor revision until the official launch.
Per the specification’s item-type notes, the Write an Email, Write for an Academic Discussion, and all Speaking tasks are AI-scored constructed responses.
Build a Sentence is also machine-scored (10 machine, 0 AI in the Writing section table), so “machine-scored” is not exclusive to Reading and Listening. This says nothing about how Reach120 scores.
Sourced from the ETS 2026 test specifications, read 2026-07-27. Everything outside this box is Reach120’s own reading of it.
What the 0-5 rubric actually rewards
Open Writing responses are scored on a 0-5 rubric. Reading that rubric as four separate demands rather than one vague impression of quality is the difference between a response that improves week to week and one that plateaus. The table below is Reach120’s working breakdown of what a strong contribution does — our reading of the task, not an ETS document.
| What the reader looks for | What earns it | What loses it |
|---|---|---|
| A clear contribution | A position stated early and held to the end. | Surveying both sides and never landing; changing sides halfway. |
| Engagement with the discussion | Answering a named peer on their own argument. | Naming a peer and then ignoring what they said. |
| Development | One reason carried through to a concrete result. | Three gestured reasons, none of them finished. |
| Language control | Varied sentence length, precise word choice, controlled grammar. | Memorised template phrasing wrapped around thin content. |
Notice what is not on that list: length, difficulty of vocabulary, and number of arguments. Long responses do not score well because they are long; they score well when the extra words did work. A 130-word post that answers a classmate and finishes one example beats a 220-word post that lists four reasons and develops none.
The two student posts are the task, not the scenery
Most published advice for this task treats the peer posts as context to be acknowledged. That is the single most expensive misreading of the 2026 Writing section. The posts are there because the task is testing whether you can enter an argument that is already running — and there are only a few honest ways to do that.
- 1Side with one and extend them. Agree with a named classmate, then add the reason or the case they did not give. Agreement without addition contributes nothing.
- 2Side with one and answer the other. State your position, then take the strongest objection from the classmate you disagree with and deal with it directly.
- 3Take a third position and use both. Reject the either/or the board has set up, and show what each classmate got right on the way to your own line.
- 4Reframe the disagreement. Argue that the two are not actually in conflict because they answer different questions, then say which question matters more.
Naming a peer versus engaging one
Weaker
I agree with Kwame. Public transport is important for cities and helps the environment. Roads are also important for people who drive.
Stronger
Kwame is right that density is the constraint — but that is an argument for putting the first bus route in the outer districts, not for postponing transit there.
The first mentions a classmate and then talks past him. The second takes his actual claim, grants it, and turns it. Only one of those is a discussion.
Conceding something real to the person you disagree with is not weakness in this task; it is the cheapest available evidence that you read the board. It costs one clause and it makes the disagreement credible.
A structure that is not a template
Memorised templates are visible, and they are visible because they arrive with content that does not fit the board. What survives is a shape rather than a script: five moves in a fixed order, each written fresh from the prompt in front of you.
- 1Position — one sentence, in the first line. Not “this is a complex issue”; the actual answer to the professor’s question.
- 2Reason — why you hold it, in general terms. One reason, chosen because you can support it.
- 3Peer — the classmate you are answering, and what you are doing with their argument.
- 4Support — one concrete example, carried to an outcome. Personal, local or hypothetical is all fine; specific is what matters.
- 5Close — the principle the argument earned. Not a summary of what you already said.
The ten minutes, spent
The Writing section is listed with an estimated 23 minutes for all 12 items, so the Academic Discussion task is the one place in Writing where a slow start cannot be recovered later. The plan below is Reach120 pacing guidance, not an ETS instruction.
| Window | What happens |
|---|---|
| First 60-90 seconds | Read the professor’s question and both posts. Decide your position before typing. A response that changes its mind halfway costs more time than it saves. |
| Next 60 seconds | Choose the one peer you will engage and the one example you will use. A second example is usually a mistake. |
| Minutes 3-8 | Write straight through in the five moves. Do not stop to fix wording; a half-finished thought is more expensive than an awkward sentence. |
| Final 60-90 seconds | Reread for verb endings, articles, and one clear position sentence. Most band-level gains actually come from this pass. |
If you consistently run out of time, the fix is almost never typing speed. It is the first ninety seconds: candidates who reread the board looking for the “right” answer are paying for a decision they could have made on the first read.
Six failure modes worth training out
- The neutral response. It explains both sides fairly and never says which one it holds. Fluent, well organised, and missing the one thing the task asks for.
- The template with a name in it. “I agree with X because this is an important issue in modern society.” The board could have been about anything.
- The four-reason list. Every reason gets one sentence, none gets a second, and the response reads as an outline of an essay nobody wrote.
- The invented statistic. “Studies show 70% of students…” — nobody is checking the number, and it still reads as filler to a reader who has seen it a thousand times. A specific personal case is more convincing and costs nothing.
- The essay in miniature. Introduction, thesis, three bodies, conclusion, compressed to 140 words. This is a discussion post; the genre is a contribution, not a composition.
- The unfinished last sentence. Running out of time mid-clause costs more than the missing words: it removes the close, which is where the strongest sentence usually sits.
How to practise this task without wasting sessions
The common failure is volume without diagnosis: twenty responses written, none reread, the same three habits repeated twenty times. A shorter, slower loop moves faster.
- 1Write one response under the real clock, with the timer visible. Untimed practice trains a task that is not on the test.
- 2Before reading any feedback, mark your own post: underline the position sentence, the peer engagement and the example. A missing underline is a diagnosis you did not need scoring for.
- 3Then read the scored feedback and fix ONE habit — the one that appears in the most responses, not the one that is easiest to fix.
- 4Rewrite the same response with that one habit corrected. Rewriting is where the change sticks; a fresh prompt lets the old habit return unchallenged.
- 5Every fifth session, write against a topic you find boring. Interest inflates practice scores and test topics do not care.
A rubric band from practice is evidence about a habit, not a predicted official score. Treat a 3 that becomes a 4 on the same prompt as the real result — that is a change in what you can do, which is what test day measures.
Academic Discussion: common questions
- How long should a TOEFL Academic Discussion response be?
- The ETS 2026 test overview, read on 2026-07-27, states that an effective response is at least 100 words and publishes no maximum. Reach120 practises to a 100-150 target because that is what fits ten minutes comfortably — it is our pacing target, not an ETS requirement or a limit, and a longer response is not penalised.
- How much time do I get for the Academic Discussion task?
- About ten minutes in practice. The 2026 specification lists an estimated 23 minutes for the whole Writing section — 12 items in total — and states that blueprint values are subject to minor revision until launch. The per-task split is Reach120 pacing guidance rather than an ETS figure.
- Do I have to agree with one of the students?
- No. A third position counts as a position, as long as the response says exactly where your line falls and why. What does not work is having no position, or having one that the peer posts played no part in.
- Can I use a template for this task?
- A fixed shape helps; fixed sentences do not. Memorised phrasing arrives with content that does not fit the board in front of you, and that mismatch is the most visible weakness in a weak response. Keep the five moves, write the words fresh.
- Is the Academic Discussion task scored by a person or by a machine?
- Per the 2026 specification’s item-type notes, Write for an Academic Discussion is an AI-scored constructed response, as are Write an Email and all Speaking tasks; Build a Sentence is machine-scored too, so machine scoring is not exclusive to Reading and Listening. That describes the official test. How Reach120 scores your practice is a separate matter, explained on our methodology page.
- What score does a good response get?
- Open responses are scored on a 0-5 rubric, and every band Reach120 shows is labelled as a Reach120 estimate. Official score reports for tests taken on or after January 21, 2026 use 1-6 bands in half-band increments, and no simple conversion between the two is published.