TOEFL Speaking, task by task.

Two: Listen and Repeat (elicited imitation across 7 sentences of increasing complexity) and Take an Interview (a 4-question simulated conversation with a video interviewer, difficulty increasing across the task). Neither existed under these names before the 2026 redesign, and neither is an integrated task — both are pure speaking, with no reading or listening summary attached.

8 minutes total across both tasks and all 11 items. ETS has not published per-task prep or response times — only that 8-minute section total is public, so treat any specific per-task timing figure you see elsewhere as unofficial.

Yes, both tasks — Listen and Repeat and Take an Interview are scored end to end by ETS’s own proprietary AI engines, with human raters stepping in on low engine confidence and a QA sample. It’s the exam’s own published scoring design, not an approximation of a human examiner.

Yes — deep coaching on both Speaking tasks ships from launch, not as a later add-on. A Verified report unlocks the full breakdown for Take an Interview and Listen and Repeat, the latter covering the whole 7-clip attempt in one unlock.

It moves from factual, personal-experience questions to ones asking you to express and support an opinion on broader issues, with difficulty increasing across the four questions — testing sustained spoken performance under rising demand, not just a single response.