The 2026 Speaking section is short — 11 responses in about eight minutes — and it asks for two very different skills. Here is how to train each one, and a daily routine that fits around everything else you are studying.
Speaking is AI-scored. ETS describes the scoring as natural language processing that assesses features including fluency, coherence and grammar. In practice that means three things matter more than anything else: you are easy to understand, you keep talking at a steady pace, and what you say is organized and accurate. Ambitious vocabulary said slowly and hesitantly usually earns less than plain, precise English said confidently.
The two tasks put weight on different parts of that. Listen and Repeat is almost entirely about accuracy and intelligibility: did you reproduce the sentence, word for word, in a way a listener can understand? Take an Interview adds content and organization: did you answer the question you were asked, and did you develop the answer? Train them separately and your weaknesses become much easier to see.
You hear a sentence once and repeat it exactly. The sentences get longer and more complex as the task goes on, and the most common way to lose credit is not mispronunciation but change: a dropped word, a missing ending, a synonym, two phrases swapped round.
Nobody can hold fifteen separate words in memory after one hearing. Everyone can hold three or four meaningful chunks. Train yourself to hear a sentence as thought groups — who or what, the action, where, when, why: The quiet room / on the third floor / closes early / on Fridays. Natural breaks fall after a long subject, before a place or time phrase, and before a connector such as because, until or if. They never fall inside on the floor or has been. Before you speak, know how many groups you heard; that number is what stops the end of the sentence from disappearing.
In fast speech, words like a, the, of, to, has, will become short and quiet. They are still in the sentence, and leaving them out changes it. The same goes for grammar endings: -s on a verb after a singular subject, -ed after has or was, plural nouns after numbers. A useful habit is to let the grammar remind you: if you heard has, the next verb almost certainly ends in a participle; if you heard three, the noun is plural. Rebuilding the small words from the grammar is more reliable than trying to remember them as sounds.
The memory of a sentence fades within seconds. Start speaking as soon as the sentence ends, say each group as one smooth unit, and keep the order you heard. If you stumble, correct the word and keep going rather than restarting the whole sentence.
How to practice: use the Listen and Repeat practice set, record every attempt, then compare your recording with the written sentence. Mark three kinds of error in different colors — missing small words, missing endings, and changed content words. Within a week, one of those colors will dominate, and that is your drill. Shadowing — speaking along with a recording a fraction of a second behind it — is a good warm-up for rhythm and linking.
You hear four questions from an interviewer, one at a time, and answer each with no planning time. The questions typically move from your own experience toward broader opinions, and the last question is often the most abstract. The biggest single improvement most test takers can make is structural: put the answer in the first sentence.
Reuse the question’s own words. If you are asked “Have you ever taught someone a new skill?”, begin “Yes, I have — last winter I taught my grandfather to make video calls.” If you are asked whether something is always a good idea, begin “Not always.” A listener (or a scoring engine) looking for relevance finds it immediately, and you have bought yourself a clear direction for the rest of the answer.
Experience questions open in the present perfect (I’ve…, I’ve never…) and switch to the past simple once you name a time (Last spring I…). Habit questions take the present simple. Imagined situations — suppose, how would you — take I’d plus a verb. Getting this right in the first five seconds makes the whole answer sound controlled.
After the answer, give one real reason (mainly because…) and one example a listener can picture: a person, a time, a number. One developed reason beats a list of four undeveloped ones. Many questions contain two jobs — for example, “What did you learn, and would you do it again?” — so count the jobs while the question is playing and make sure the second one gets answered, then finish with a short line that returns to your answer.
Silence costs more than an imperfect phrase. Have a few ways to keep going: a short opener that leads straight into your frame (“Well, I’d say…”); a quick self-correction with no restart (“on Friday — sorry, Saturday”); a description when a word is missing (“the thing you use to…”). One or two fillers per answer is natural; a chain of them is not.
How to practice: work through the interview practice questions in sets of four, timed and recorded, so you rehearse giving the last answer the same energy as the first. Then review each recording against the TOEFL speaking rubric.
Twenty to thirty minutes a day is enough to make steady progress, and short daily sessions beat one long weekly session for fluency. Record everything; the improvement comes from listening back, not from speaking more.
Speaking is one of four equally weighted sections, so see how a higher Speaking band moves your overall score with the score calculator, and fit this routine into a wider TOEFL study plan. For the format itself, see the Speaking section guide.
Practice both Speaking tasks
Short lessons on thought groups, endings and interview answers, with recorded practice.
Official pages checked for this guide. Last reviewed October 2026.