Interactive Speaking

It is a spoken-conversation task added in July 2025. An animated character asks a series of about six to seven questions, one at a time, and you record a spoken answer of up to roughly thirty-five seconds to each. Every question is generated from your previous answer, so it works like a real, adaptive conversation rather than a set of fixed prompts.

1 July 2025. It replaced the Listen, Then Speak task as part of a wider update that also removed Read Aloud and expanded Interactive Listening. If a guide still lists Read Aloud or Listen, Then Speak as current tasks, it has not been updated for the current format.

Up to about thirty-five seconds per answer. There is a brief moment to gather your thoughts after each question, then a recording window. You need not use the whole time, but very short answers give the scorer little language to assess.

Around six to seven linked questions in a single conversation. The exact number can vary because the test adapts, but it is a short, connected exchange rather than one isolated answer.

No. Each question plays once, with no transcript and no replay, so listening accurately the first time is part of the task. Listen for the question word and the key detail rather than trying to hold every word.

It feeds your Speaking subscore, and each answer is rated on relevance, fluency, grammar, vocabulary and how clearly you can be understood. The exchange is read as one connected conversation, and the single most important factor is whether each answer actually responds to the question asked.

Yes. Interactive Speaking replaced Listen, Then Speak in the July 2025 update. The old task asked one question and took one answer; the new one is a multi-question conversation where each question builds on your last reply.

Not fixed answers. Because each question is generated from what you just said, prepared content usually will not fit the question you are actually asked, and off-topic answers lose relevance marks. What you can prepare is a shape: answer the question directly in one sentence, then add one sentence of reason or example.

A specific one. A concrete first answer — a named hobby, a real reason, a particular example — gives the follow-up questions something to build on and gives you material to keep talking about. A vague first answer forces generic follow-ups and leaves you starting from nothing each time.

Answer the part you did catch rather than narrating confusion. Because each question plays once, a partial but relevant answer scores better than silence or "sorry, I did not understand." Listen especially for the question word, which tells you what kind of answer is wanted.

It is one of the more demanding tasks, because it combines listening on one hearing with speaking on the spot and you cannot rely on prepared material. But it is very trainable: the two-sentence answer shape and practising short question chains cover most of what it tests.

It feeds Speaking directly, and through the DET's integrated subscores it also contributes to Conversation and Production. Because you are responding to spoken questions, your listening accuracy affects your answers too.

The Speaking Sample is one long, uninterrupted response to a single open prompt, recorded for institutions to review. Interactive Speaking is a back-and-forth of several short answers where each question depends on your last one. The Sample tests sustained monologue; Interactive Speaking tests responding in a live exchange.

Drill a two-sentence answer to any question — direct answer, then one reason or example — until it is automatic, practise answering on a single hearing with no replays, and run short chains of follow-up questions so you get used to each one building on your last answer. Recording yourself and filling the full time with real content is the core exercise.

A working microphone, yes — it is a spoken task. A webcam is required for the test overall (proctoring and the video the Speaking Sample produces), but the answers you give in Interactive Speaking are audio. Run the equipment check before you start so the microphone is confirmed.