Three people discuss an academic topic. You hear it once, you never see the words, and then you have two minutes to summarise what each of them said and where they agreed. It's one of the two newest PTE Academic tasks — and the one where the popular advice is most likely to actively hurt you.
Audio only, once, no transcript. That single fact drives everything else.
You hear three people discussing an academic topic. Pearson's preparation material is blunt about the format: "You will not see this when listening." There's no transcript, no subtitles, and the audio plays once only. Then a 10-second preparation countdown runs, the microphone opens, and you have two minutes.
Because you can't re-read anything, the task is really testing whether you can hold three people's positions in your head at once — and that's why Pearson explicitly tells you to take notes while listening. Notes aren't a workaround here; they're the intended method.
Speaker 1 opens with a position — "Schools should encourage students to learn a second language; it improves communication and career options."
Speaker 2 partly agrees but adds a caveat — "It's valuable, but real proficiency takes time schools may not have."
Speaker 3 shifts the angle — "Technology and flexible approaches could close that gap."
That shape — position → caveat → new angle — is typical. The speakers rarely fight, and they rarely reach a neat conclusion. Which matters more than it sounds, as you'll see below.
Three traits, and they're not weighted equally.
The biggest single trait. Did you cover each speaker's main point, in your own words, and show how their views relate? Partial credit applies.
Smooth, natural, even delivery across the full response — no restarts, no long hunting pauses.
Each word clearly produced. Accent isn't penalised; unclear sounds are.
Two things follow from that 0–6 on Content. First, Content is where the task is won — it carries more range than either speech trait. Second, and this is the part almost nobody mentions: a human expert reviews the Content of this task before your score is finalised. Pearson applies that human review to a specific set of question types, and Summarize Group Discussion is one of them. Pronunciation and fluency are machine-scored; your content gets read by a person.
That's a good reason to stop trying to game a template past a scoring algorithm. A person is going to look at it.
Sources: Pearson's test format — Speaking & Writing, the official task article, and the PTE Academic Test Taker Score Guide (July 2025).
This is the part where the popular advice and Pearson's own advice point in opposite directions.
Search this task and you'll be given a fill-in-the-blanks template: "The discussion focused on… The first speaker argued that… The second speaker mentioned… They concluded that…"
Pearson's published guidance for this exact task lists, under things not to do, introducing each topic in a mechanical way — topic one is, topic two is, topic three is. That is precisely what those templates produce.
It gets sharper. Pearson classifies pre-prepared or memorised material as an irrelevant response — and an irrelevant response scores zero on Content. On a task where Content is the 0–6 trait and a human reads it, a memorised frame is a genuine risk, not a safe fallback.
Nearly every template ends with "They concluded that…" or "No consensus was reached." But Pearson never asks for a conclusion. What it asks for is that you say where the speakers agree or disagree — which is a different thing. Real discussions often end without concluding anything, and forcing a fake conclusion onto one either invents content that wasn't there (which costs Content) or burns your last seconds saying nothing.
The second one names the speakers, tracks a point of tension, and shows Lucas responding to Isabella. That's the relationship-between-viewpoints thing Pearson's top band is asking for. It's also not a template — it's just what you'd say if you'd genuinely listened.
Structural guidance, not a script. There's a real difference.
And the published don'ts: don't simply list who said what in running order, don't add information that wasn't there, don't confuse the speakers, don't try to mention every single point, don't jump between ideas — and don't introduce topics mechanically.
Covered above, and it's first for a reason: Pearson treats pre-prepared material as an irrelevant response, and Content is the trait with the most range.
✓ Fix: learn the ingredients (each position, the agreement/disagreement, how they relate) rather than the sentences. Same preparation benefit, none of the risk."Speaker one said X. Speaker two said Y. Speaker three said Z." It's accurate and it still scores in the middle, because the top of the Content band wants the relationships between views, not an inventory.
✓ Fix: find the one disagreement or the one build. Structure the answer around that tension instead of around the speaking order.Re-tell Lecture has one voice and one argument, and 40 seconds. This task has three voices, competing views, and three times the talking time. Advice ported straight from Re-tell Lecture leaves you a minute short and structurally flat.
✓ Fix: practise this task on its own timing. Two minutes is much longer than it sounds when you're standing there.The most common panic. You summarise the three positions, you're done at 45 seconds, and there's over a minute of silence left — which reads as hesitation and pulls your fluency score down with it.
✓ Fix: the second half is where supporting detail and the agree/disagree analysis go. Say what each speaker's reason was, not just their position — that's both more content and more time.Attributing Isabella's caveat to Nathan is a content error, and an easy one to make from memory alone when three voices blur together.
✓ Fix: this is exactly what the notes are for. One column per speaker, names at the top, from the first second of audio.Cramming every detail is on Pearson's published don't list, and it forces you to rush — which costs fluency and pronunciation while adding nothing to content.
✓ Fix: main point plus one reason per speaker, then the relationship between them. Two minutes is enough for that at a calm pace.The hard part is listening to three voices without a transcript. Practising with the transcript visible trains the wrong skill entirely.
This is where a lot of practice goes wrong. Reading the discussion while you listen feels productive, but the exam gives you audio only, once — so reading along rehearses a task you'll never sit. It's the same reason people who "practise" with subtitles are still lost on test day.
On The PTE Master, Summarize Group Discussion runs the way the exam does:
Three voices, audio only, notes pad, two minutes on the clock — the way the real task actually runs.
Start practising