Summarize Group Discussion(SGD)
19% of SPEAKING20% of LISTENINGAudio up to 3 min · 10s prep · 2 min speaking
NEW in Aug 2025 — and already the biggest Listening contributor (20%) plus 19% of Speaking.
Research-seeded guide — verified method and traps included; a full live example is coming next.
The Format
- Audio
- Group discussion, usually 3 speakers, up to ~3 minutes
- Speak
- 2 minutes — the longest speaking response in the test
The Scoring
0–6
Content
Grasp of main ideas and each speaker's viewpoint. Human-reviewed.
0–5
Oral Fluency
0–5
Pronunciation
~19% of Speaking + ~20% of Listening — the highest-paying new task. Robotic template answers can score ZERO.
The System
Tag by Speaker
1
TAG
S1 / S2 / S3 + keyword
Note each speaker's opinion in 3–4 second listening/writing bursts. Don't plan during audio.
2
MAP
Agreements vs disagreements
Who agreed, who pushed back, on what.
3
SUMMARIZE
Topic → each view → wrap-up
"The discussion was about [topic]. The first speaker felt ___… However, the second disagreed, arguing ___… Overall, they agreed ___." Natural wording — a human is listening.
Avoid
The 3 Traps That Cost You the Task
✕ Robotic template delivery
Human raters can zero templated responses. Frameworks yes, scripts no.
✕ Losing a speaker
Missing one speaker's view caps content.
✕ Fizzling early
Speak steadily for most of the 2 minutes.
Daily Drills
- →Practice with panel podcasts: tag 3 speakers' views, then 2-min spoken summary.
Memory Hooks
TAG → MAP → SUMMARIZEOne view per speaker, minimum