How do I prepare for TOEFL Write for an Academic Discussion?

What does the Write for an Academic Discussion task ask you to do?

Write for an Academic Discussion asks you to state your opinion on an online class discussion board and give reasons for it, and you have 10 minutes to write (Test Overview page 15, checked 2026-10-30). A post from the professor briefly introduces the topic and asks a question seeking opinions, and two students then post short responses that take different positions. The test taker adds their own position and supports that opinion with their own reasoning, experience and knowledge (same page). The Writing section has 1 item of this task, worth up to 5 points (ETS "TOEFL iBT Test: 2026 Update Test Blueprint and Specifications Document", hereafter the Specifications Document, checked 2026-10-30).

The directions for the sample task in the Test Overview ask you to state and support your opinion and to contribute to the discussion in your own words, and they say an effective response has at least 100 words (Test Overview page 15). The ETS guide to the Writing section explains that this task measures the ability to develop ideas, to respond to other people's perspectives and to write in an academic tone (ETS "TOEFL iBT Writing Section", checked 2026-10-30). The Specifications Document classifies this task as AI-scored (a method that uses natural language processing and machine learning to rate features such as fluency, coherence and grammar). The ETS "TOEFL iBT Technical Manual" (hereafter the Technical Manual, ETS Research Report RR-25-12, October 2025, checked 2026-10-30) says this task is scored with both the ETS AI scoring engine and human raters, and that human raters give the score for responses the AI is not confident about or finds difficult (Technical Manual pages 35 and 39, page numbers follow the PDF page order).

All scores in this article are on the 0 to 5 point scale for a single task. This is a different scale from the Writing 1 to 6 bands on the score report, and how to read the bands is covered in the Unlock Prep (unlockprep.com) guide "How do TOEFL 1 to 6 scores compare with 0 to 120 scores?". To see all three Writing tasks at once, read "How do I prepare for the three new TOEFL Writing tasks?" first.

How does the ETS scoring guide divide 0 to 5 points?

The ETS scoring guide gives each score a name and a heading sentence, then lists the features seen in a typical response at that score. The table below is our short paraphrase of the ETS wording (Test Overview pages 24 to 25; the same wording appears on page 2 of the ETS "Writing Scoring Guide", checked 2026-10-30).

The 1-point level has one more line: a response with little language of its own, where most of what makes sense is borrowed from the texts in the task (the stimulus, which we read as the professor's post and the students' posts). A score of 0 goes to a blank response, a response that rejects the topic, a response not in English, a response that copies the prompt in full, a response unrelated to the prompt, or a response of random keystrokes (same scoring guide).

Compared with the Write an Email scoring guide, this one has no line on social conventions (politeness, level of formality). Instead, every heading sentence includes the phrase "contribution to the online discussion". The Write an Email lines are covered separately in the companion article "Where do you lose points on TOEFL Write an Email?".

Which sentences separate the two answers we wrote?

The two answers part ways in five places: the supporting sentence, explanation and example, responding to another student's opinion, sentence forms, and the closing sentence (our view). The task and answers below are all examples we wrote for illustration. We wrote the task, and the score levels are not ETS scoring results but our judgment from holding the answers up against the scoring guide wording.

Task we wrote (business class discussion):

Professor Hall: Some companies have tested a four-day work week, in which employees work four days instead of five for the same pay. Supporters say rested workers are more productive. Critics worry that customers and teams will suffer. Should more companies adopt a four-day work week? Why or why not? Mina: Yes. When people get a three-day weekend, they come back with more energy, so they get more done in four days than they did in five. Daniel: I disagree. Many businesses, like restaurants and clinics, need staff every day. If everyone takes Friday off, customers will have to wait longer.

Answer A we wrote (Our view: close to the 3-point level, 100 words)

I agree with Mina. I think four-day work week is good idea for most companies. When people have three-day weekend, they come back with more energy. So they can do more work in four days. Also, workers can spend more time with family and they will be happy. Happy workers work hard. Many people want more free time these days, so it is very important. Daniel said customers will wait longer, but I think it is not big problem. In conclusion, companies should adopt four-day work week because it is good for workers and it is good for companies too.

Answer B we wrote (Our view: close to the 4-point level, 117 words)

I partly agree with Daniel, but I still think most office-based companies should try a four-day week. Daniel is right that a clinic cannot simply close on Fridays. However, the professor asks about most companies, and many of them does not serve customers in person every day. For these firms, a shorter week can push teams to cut meetings that waste time. For example, when my cousin's design company tested a four-day schedule, the staff agreed to hold only one meeting a week, and they still finished their projects on time. Businesses that must stay open could let employees take different days off. In this way, companies could keep the benefits Mina describes without making customers wait.

Both answers reach the 100 words in the ETS sample directions, so length does not separate them. Here is how they compare in each of the five places.

1. Supporting sentence. A "When people have three-day weekend, they come back with more energy. So they can do more work in four days." → B "For these firms, a shorter week can push teams to cut meetings that waste time." A's reason copies Mina's post almost word for word and does not meet the directions' request to contribute "in your own words". B only points to Mina's reason at the end and gives a new reason of its own. Lines involved: the 1-point line "most of what makes sense is borrowed from the texts in the task" (only these two sentences of A touch this line), and the 4-point heading sentence "relevant contribution".
2. Explanation and example. A "Also, workers can spend more time with family and they will be happy. Happy workers work hard." → B "For example, when my cousin's design company tested a four-day schedule, the staff agreed to hold only one meeting a week, and they still finished their projects on time." A jumps from a reason to a conclusion and gives no example. Lines involved: the 3-point line "some explanations, examples or details missing or unclear", and the 4-point line "adequately developed explanations, examples and details".
3. Responding to another student's opinion. A "Daniel said customers will wait longer, but I think it is not big problem." → B "Daniel is right that a clinic cannot simply close on Fridays. However, the professor asks about most companies, …" A brushes the opposing view aside without a reason, while B concedes the point and then answers it with the scope of the professor's question (most companies). The scoring guide has no separate line for responding to other opinions; this ability is described in the task description on page 15 of the Test Overview and in the ETS guide to the Writing section. The line it touches in the scoring guide: the 3-point line "some of the explanation missing or unclear".
4. Sentence forms and vocabulary. A strings together short sentences that open with "I think", and drops the article before "good idea", "three-day weekend" and "big problem". B mixes sentences linked with "However", relative clauses (that waste time, that must stay open) and forms such as "Daniel is right that …". Lines involved: the 3-point lines "some variety of structures" and "noticeable errors", and the 4-point lines "variety of structures, appropriate words" and "few errors".
5. Closing sentence. A "In conclusion, companies should adopt four-day work week because it is good for workers and it is good for companies too." → B "In this way, companies could keep the benefits Mina describes without making customers wait." A repeats its first sentence, while B ties the two students' views together in one sentence and adds to the discussion. Our view is that the line involved is "contribution to the discussion" in the heading sentence.

Where B falls short of the 5-point level is its second reason (our view). "let employees take different days off" ends in a single sentence, with no explanation or example attached. The 5-point level describes "well developed" explanations, examples and details, and the 4-point level describes "adequately developed" ones. We also left one subject-verb agreement error in B (many of them does). However, the sample response in the Test Overview received 5 points, the only error the ETS commentary pointed out was one article, and the commentary describes that error as one that could appear even in a first draft by the most proficient writers (Test Overview page 17, checked 2026-10-30). So we measure the distance between B and the 5-point level by the depth of explanation rather than the number of errors.

This article does not show a separate 5-point answer. In our view, the professor's question and the two student posts change with every task, so it is hard to use any one answer as a model.

What should I practice first?

Practice by matching one exercise to each line of the scoring guide (our view). ETS states only the 10-minute time limit and does not say how to divide the time within it.

1. Write your position, answering the professor's question, in the first sentence. Bringing one key word from the question (in the task above, most companies) into your position sentence lowers the risk of an irrelevant answer.
2. Attach one example or detail to each reason. As in the second difference above, find sentences that stop at "so they will be happy" and add one more sentence on what, when or how.
3. Respond to one of the two student posts, but answer in your own words instead of copying that student's sentences (the first difference).
4. Write to the end within 10 minutes, then use any time left to check errors such as articles and subject-verb agreement. Test Overview pages 14 and 15 say both writing tasks are rated even on mechanical conventions such as spelling, punctuation and capitalization.

Can I memorize a template?

"You can use a framework, but the wording should be your own. If memorized sentences do not fit the task, you may lose points." That is Unlock Prep's position, and industry guides are split between supporting templates, allowing a framework only, and leaving it optional.

The ETS documents touch this position in three places. The directions for the sample task in the Test Overview say to contribute to the discussion "in your own words" (page 15). The 1-point level describes a response where most of what makes sense is borrowed from the texts in the task, and the 0-point level describes a response that copies the prompt in full (pages 24 to 25). In our view, if a memorized opening sentence that fits any topic does not answer the professor's question, the response moves away from "relevant contribution" in the heading sentences of the 5-point and 4-point levels. Table 4 on page 37 of the Technical Manual lists "similarity to question prompt" among the example features the scoring engine looks at for this task. It does not say in which direction that feature moves the score.

What has ETS not disclosed?

In the Test Overview, the scoring guide, the Specifications Document, the Technical Manual, the ETS information bulletin (2026-27) and the Writing and Test Content pages on ets.org, we either did not find the four items below or found only the method without numbers (checked 2026-10-30).

• How much the score changes when a response falls short of 100 words. The line saying an effective response has at least 100 words appears in one place, the sample task in the Test Overview, and nothing says whether every item on the real test carries the same line. The scoring guide does not set scores by word count.
• Whether you must respond to the student posts. The task descriptions on Test Overview page 15 and Technical Manual page 24 join responding to other claims and using information from the short posts with "and/or", and the scoring guide has no such line.
• A conversion table from task scores to Writing bands. The method is disclosed. Table 2 on page 30 of the Technical Manual gives the Writing raw score as 0 to 20, the sum of Build a Sentence 0 to 10, Write an Email 0 to 5 and this task 0 to 5, and page 35 says this total is converted to a scaled score for each test form by weighted equipercentile linking (a statistical procedure that adjusts for small differences in difficulty between test forms). The conversion tables are not published.
• How much weight AI scoring gives each line of the scoring guide. Table 4 on page 37 of the Technical Manual lists example features for this task (number of sentences, discourse coherence, similarity to the prompt, sentence variety, word frequency, collocation accuracy), but it does not give the weights.