Sample answers · Summarize Group Discussion
PTE Summarize Group Discussion · Speaking + Listening
New Aug 2025PTE Summarize Group Discussion sample – Band 79.
A worked SGD: a 2 minute 24 second 3-speaker academic discussion rendered as transcript, the 3-column speaker grid a Band 79 candidate captures during playback, a 258-word 2-minute spoken summary, and a trait breakdown across all three scored traits. SGD was added in the August 2025 format update and is one of 7 task types where Content is marked by both an AI model and a human reviewer.
Last verified 17 July 2026 · Written for PTE Academic post-August 2025 format · Verified against Pearson's July 2025 Score Guide.
The stimulus discussion
Should university lectures be recorded and made available online?
2 minutes 24 seconds audio · 3 speakers · plays once, no replay control · accompanied by a discussion graphic · no on-screen transcript on the real exam.
Companion image on screen: Three academics around a small round table with microphones, a laptop and notepads, mid-discussion. A common on-screen visual for Summarize Group Discussion.
PROFESSOR CHEN: Thanks everyone for joining. The question we're tackling today is whether university lectures should be recorded and made available online after the class. My view is that recording should be standard practice. Recorded lectures let students revisit difficult concepts, they support students whose first language isn't English, and they're essential for students with disabilities or who fall ill during term. DOCTOR MARSH: I disagree, and I want to push back on this. When I've recorded my lectures in the past, attendance dropped by around forty percent within a fortnight. Students who watch at double-speed at midnight the night before an exam absorb far less than students who show up, participate, and discuss the material live. We're also seeing recordings get shared on messaging apps and paid tutoring sites without our consent. DOCTOR ALVES: I think both of you are partially right. The accessibility case is compelling. I don't think we can ethically deny recordings to a deaf student or a student who's been hospitalised. But Marsh has a point about engagement. My compromise would be this: record every lecture, but restrict access to enrolled students through the university portal only. Add a watermark. And crucially, make live participation part of the assessment so students still have a reason to attend live. PROFESSOR CHEN: The watermark idea is smart. The attendance-as-assessment part I'd push back on though. Mandatory attendance treats undergraduates like schoolchildren. Adults should choose how they learn. DOCTOR MARSH: But that's precisely the problem. Twenty percent of my third-year students last semester never attended a single live class. They watched every recording, scored fine on the exam, and left with a degree, but they never developed the skill of processing information in real time. That's what we owe them. DOCTOR ALVES: Perhaps the answer is discipline-specific. In lecture-heavy subjects like mine, recording is fine. In seminar-driven subjects, live attendance matters more. A blanket rule for a whole university doesn't fit.
The transcript is shown for study purposes only. On the real PTE Academic exam, the audio plays once and there is no visible text. Your summary is built entirely from what you heard and captured in notes.
Speaker grid captured during playback
3 columns, one per speaker, shorthand only.
| Speaker | Position | Reasons | Evidence / detail |
|---|---|---|---|
| Prof Chen | FOR universal recording | Access + revision, ESL support, disability + illness accommodations | Pushback later against mandatory attendance ("adults should choose") |
| Dr Marsh | AGAINST recording | Attendance drops → weaker engagement → shallower learning. Unauthorised sharing risk | Attendance ↓ 40% in 2 weeks in her own courses. 20% of Y3 students never attended live last semester |
| Dr Alves | COMPROMISE | Record but: portal-only access, watermark, live-participation as graded assessment | Discipline-specific: recording fine for lecture-heavy, live-attendance for seminar-driven |
Grid drawn on the erasable whiteboard during the on-screen instruction, BEFORE the audio starts. Filled in shorthand during the 2:24 of playback. Every time a new voice begins, move your pen to that speaker's column – attribution is the SGD-specific Content signal.
Delivery version · Band 79
258-word summary in 103 seconds.
258 words · 103 seconds at 150 wpm · fills the 2-minute recording window with 17 seconds of headroom, giving you room to slow down on the closing paragraph. Four paragraphs: one per speaker plus one on the disagreement pattern the panel closes on.
Paragraph-by-paragraph
How each speaker cell became one paragraph.
Paragraph 1 (topic + Chen's position, 49 words)
Names the topic in the first sentence and gives Chen's three reasons in the second. Ordered by speaker, not by argument type – SGD summaries are easier to follow when they trace the discussion as it unfolded.
Paragraph 2 (Marsh's counter-argument, 60 words)
Preserves Marsh's specific quantitative claim ("forty percent within two weeks") which is the load-bearing evidence in her position. Human reviewers look for at least one specific numeric claim from the discussion.
Paragraph 3 (Alves's compromise, 64 words)
Alves's compromise is the most substantive position. Enumerating her three concrete proposals (portal-only, watermark, graded participation) shows Content depth and demonstrates syntactic control across parallel clauses.
Paragraph 4 (the disagreement pattern, 85 words)
Captures the pattern of the discussion itself – not just what each person said in isolation, but how they respond to each other and where they land. This is the SGD-specific skill that separates it from Retell Lecture.
Trait breakdown
Band 79What earned each trait score.
| Trait | Score | Why it scored there |
|---|---|---|
| Content | 5/5 | All three speakers' positions preserved with the specific evidence they cited (forty percent attendance drop, twenty percent no-shows, portal-only + watermark + graded participation). The disagreement pattern is captured in the closing paragraph, not just each position in isolation. Content on SGD is one of the 7 hybrid-scored traits: AI plus a human reviewer both mark it. |
| Oral Fluency | 4/5 | Steady pace at 150 wpm across the whole 103-second delivery. Signposted transitions between paragraphs ("Doctor Marsh disagrees", "Doctor Alves proposes", "The disagreement then narrows"). Deduct 1 for one hesitation while switching between Alves and Chen mid-paragraph 4. |
| Pronunciation | 4/5 | Clear articulation on all speaker names and technical vocabulary (accommodations, unauthorised, discipline-specific). Deduct 1 for slightly compressed stress on "ethically compelling" – a Band 90 delivery would space those syllables more evenly. |
| Total | 13/15 | Consistent with Band 79 across both Speaking and Listening (SGD feeds both skills). |
Note: SGD Content is scored by AI PLUS a human expert reviewer (one of 7 hybrid-scored task types on the PTE Academic exam). If AI and human disagree on Content, a second human decides. Oral Fluency and Pronunciation are AI-only.
Timing plan
2:24 audio, 10s prep, 2 min speak.
| Time | Stage | What to do |
|---|---|---|
| 0s | Audio + image appear | The up-to-3-minute discussion audio starts. A discussion/people graphic sits alongside. Draw your 3-column speaker grid on the whiteboard in the first 5 to 8 seconds. |
| 0 to 2:24 | Capture the speaker grid | One column per named speaker. Rows: POSITION, REASONS, EVIDENCE. Fill in shorthand as each speaker takes a turn. When a speaker changes, move your pen to that column. |
| 2:24 | Audio ends | 10-second preparation window opens. Look at your grid. Decide on your opening sentence and the order you'll walk through the speakers. |
| 2:34 | Beep + recording starts | There IS a short beep on SGD (like Retell Lecture). Begin speaking within 1 second. You have 2 full minutes to deliver – you do not need to rush. |
| 2:34 to 4:34 | Deliver the 2-min summary | Aim for 260 to 300 words at 130 to 150 wpm. One paragraph per speaker, plus a closing paragraph on the disagreement pattern. Continuous flow between paragraphs, no long pauses. |
| ≥ 3s silence | Auto-stop | Recording auto-stops after 3 seconds of silence, and hard-stops at the 2-minute limit. Finish your closing sentence and hold. |
The 3-column speaker method
5 steps for turning a 3-minute discussion into a 2-minute Band 79 summary.
Draw the 3-column grid BEFORE the audio starts
Three columns, one per speaker slot (Speaker 1, Speaker 2, Speaker 3). Rows: POSITION, REASONS, EVIDENCE. Total setup time: 8 to 10 seconds. Do this during the on-screen instruction, not during the audio.
Update the SPEAKER LABEL the moment a new voice starts
SGD stimulus audio always names each speaker at least once. Write the actual name (Chen, Marsh, Alves) in your column header the first time you hear it. Speaker attribution is a specific Content signal on SGD.
Capture at least ONE specific number or claim per speaker
SGD discussions include quantitative evidence ("forty percent", "twenty percent", "three billion pounds"). These are what separates a Band 79 summary from a Band 65 gist. Copy them verbatim into the EVIDENCE row.
Use the 10s prep window to plan structure only
Ten seconds is not long enough to rehearse a 2-minute summary. Use it to decide (a) your opening sentence, (b) which speaker you'll open with, and (c) whether you'll close with a disagreement paragraph. The rest is improvised from the grid.
One paragraph per speaker + one closing paragraph on the pattern
Structure your 2-minute delivery as 4 paragraphs: 3 speaker paragraphs (in the order they appeared or in position order) + 1 closing paragraph capturing what they agreed and disagreed on. That structure fills the 260 to 300 word target naturally.
6 common SGD mistakes
The failure modes that drag a Band 79 to a Band 60.
| Mistake | What it costs you |
|---|---|
| Attributing arguments to the wrong speaker | Speaker attribution is a Content signal on SGD (unlike Retell Lecture). Confusing Chen's argument with Marsh's or Alves's costs Content marks even if the underlying points are all captured. |
| Summarising only one or two of the three speakers | Content on SGD requires all three positions to be captured. Missing one speaker entirely usually caps Content at 2 or 3 out of 5, regardless of how well the other two are handled. |
| Delivering a 60-second summary instead of using the full 2 minutes | SGD has a 2-minute window. Under 90 seconds usually means you did not have enough Content or paragraph structure. Content and Oral Fluency both drop when the summary is too thin for the window. |
| Trying to include every specific example or quote from the discussion | 3-minute discussions include far too much detail for a 2-minute summary. Pick the load-bearing evidence per speaker (usually one number, one example). Everything else is compression noise. |
| Ignoring how the speakers respond to each other | SGD-specific skill: the discussion pattern (agreement, disagreement, compromise) is part of Content. A summary that reports each speaker in isolation without noting the interactions misses the discussion structure. |
| Reading a stock template opening ("This discussion was about...") | AI models detect stock openers across candidates. Vary opener across practice: "The three academics debate...", "The panel discusses...", "The three speakers consider...". Structure stays the same, opener varies. |
FAQ
Summarize Group Discussion, answered.
How is PTE Summarize Group Discussion scored?
Three traits per the July 2025 PTE Academic Score Guide: Content (0 to 5), Oral Fluency (0 to 5) and Pronunciation (0 to 5). Content is hybrid-scored, meaning both an AI model and a human expert reviewer mark it – SGD is one of only 7 task types with human-reviewed Content. Oral Fluency and Pronunciation are AI-only. Maximum raw score per item is 15.
How many Summarize Group Discussion items are on the PTE Academic test?
2 to 3 items per test per Pearson's July 2025 Score Guide. SGD was added on 7 August 2025 as one of the two new Speaking task types (alongside Respond to a Situation), taking the total number of PTE Academic task types from 20 to 22.
How long is the audio for Summarize Group Discussion?
Up to 3 minutes per Pearson's Score Guide, featuring three named speakers in an academic panel format. The audio plays exactly once with no replay control, alongside an on-screen discussion/people graphic. You then have 10 seconds of preparation before the 2-minute recording begins.
How long should my Summarize Group Discussion answer be?
Aim for 260 to 300 words, delivered in 105 to 115 seconds at 140 to 150 words per minute. That fills the 2-minute recording window and leaves 5 to 10 seconds of buffer. Delivering under 90 seconds usually means the summary was too thin for the window; over 2 minutes will be cut off mid-sentence.
Do I need to remember each speaker's name?
It helps but is not strictly required. Naming speakers by their announced label (Professor Chen, Doctor Marsh, Doctor Alves) is stronger Content evidence than generic labels ("the first speaker", "the woman"). If you cannot remember a name, using role-based labels ("the professor", "the researcher") still scores – but attribution matters, so at minimum keep the three positions distinct.
What is the difference between Summarize Group Discussion and Retell Lecture?
Both are hybrid-scored Speaking + Listening tasks with the same three traits, but the stimulus and output are different. Retell Lecture is a 90-second single-speaker lecture, with 10 seconds prep and 40 seconds to speak. SGD is up to 3 minutes with three speakers, with 10 seconds prep and 2 minutes to speak. SGD tests your ability to track and attribute multiple positions; Retell Lecture tests your ability to compress a single argument.
Keep going with Summarize Group Discussion
One pattern, three depths.
You've just read the band 79 worked example. Here's the rest of the loop for this task.
Reusable framework
Summarize Group Discussion template
The reusable answer framework — structure, sentence starters, timing marks — that fits any Summarize Group Discussion prompt.
Read the template →Interactive drill
Practise Summarize Group Discussion
Free interactive drills — do the reps yourself with the model answer revealed after each attempt.
Start the drill →Further reading
Related tools.
- → Retell Lecture sample – the sibling single-speaker task with a 40-second spoken output.
- → Summarize Spoken Text sample – the sibling written listening task.
- → Speaking section overview – all speaking task types on PTE Academic.
- → PTE 2025 format changes – full context on the August 2025 update.
- → Take a free scored mock – SGD items with AI scoring.