TOEFL 2026 Speaking – Listen and Repeat and Take an Interview

Imagine this: you’ve spent three weeks cramming TOEFL Speaking response templates. “According to the reading passage… The professor disagrees because…” – you know these structures by heart. You walk into the exam, and instead of four tasks involving reading, listening, and planning, you’re given seven sentences to repeat and four questions you must answer immediately, with zero preparation time. Templates? Useless. Lecture notes? There’s no lecture. There’s only a microphone, your voice, and an AI algorithm evaluating every syllable.

This is TOEFL 2026 Speaking – the section that has undergone the most radical transformation in the exam’s history. ETS has removed all four old tasks and replaced them with two entirely new ones: Listen and Repeat (repeat the sentence you hear) and Take an Interview (answer questions without preparation). The duration has dropped from 17 to 8 minutes. You’re no longer assessed by humans, but by an AI algorithm trained on millions of recordings. And pronunciation nuances, such as ‘th’ sounds often pronounced as ‘f’ or ‘t’, or vowel lengths that remain constant regardless of context, suddenly become far more critical than ever before.

In this guide, we break down the new Speaking section. You’ll learn exactly how both tasks work, what the AI assesses, common pronunciation errors made by non-native speakers, and how to prepare when you can no longer rely on memorized templates. If you’re planning to take the TOEFL 2026, this section demands an entirely new approach to practice.

Listen and Repeat: A Detailed Breakdown

Let’s start with a task that has no equivalent in any other popular language exam. Listen and Repeat consists of 7 sentences you’ll hear through headphones, one after another, and you must repeat each one as accurately as possible. You won’t see the text on screen. You have no preparation time. You hear a sentence, immediately repeat it, and move on to the next.

Sounds simple? At first glance, yes. But ETS didn’t add this task for you to relax; they added it to test something the old TOEFL never directly assessed: your pronunciation, intonation, and phonetic fluency. In the old format, you could have average pronunciation and still score 25+/30 because examiners focused on content and organization. Now, the algorithm compares your recording to a model, and differences in sounds, word stress, or sentence melody directly impact your score.

What exactly does this task involve?

You sit in front of a computer with headphones and a microphone. A playback icon appears on the screen. You hear a recording of a native speaker – a single sentence on topics related to campus life, everyday situations, or general subjects. After the sentence ends, you have a few seconds to repeat it into the microphone. Then, the next sentence plays automatically.

Key mechanic: sentences become progressively more difficult. The first sentence might be: “The library closes at ten o’clock.” – short, simple, with clear intonation. The seventh sentence will be something like: “The professor suggested that students who hadn’t completed the prerequisite should consider enrolling in the introductory seminar before attempting the advanced course.” – longer, with several subordinate clauses, complex word stress patterns, and natural vowel reduction.

What exactly does the AI algorithm assess?

The ETS algorithm analyzes your recording across four dimensions:

Pronunciation accuracy: whether the individual sounds you produce match standard English sounds. This doesn’t mean you need to sound like an American from Kansas, but your “th” must be recognizable as /θ/ or /ð/, not as /f/ or /d/. Your vowels must be distinguishable – “ship” cannot sound the same as “sheep”.

Prosody: whether your sentence has a natural English melody. English is a stress-timed language, where key words are emphasized, and function words (the, a, to, of) are reduced. Many non-native languages have a more “flat” intonation and consistent word stress, which can make speech sound “foreign” to native speakers.

Fluency: whether you speak without hesitation, pauses, false starts, or self-corrections. The algorithm measures the time between the end of the original recording and the beginning of your response, the length of pauses within your speech, and your overall speaking pace. The goal isn’t to speak quickly, but to speak smoothly and consistently.

Listening accuracy: whether you repeat the exact words you heard. If the sentence is “She’s been studying for the exam since Monday”, and you say “She has been studying for the exam since Monday” – the algorithm will detect this. Contracted forms (she’s, I’ve, they’d) are part of natural English, and “expanding” them to full forms signals that your ear isn’t catching natural speech patterns.

Why is this task a challenge for non-native speakers?

Let’s be honest, pronunciation is a common challenge for many non-native English speakers. This isn’t due to laziness, but because many native phonetic systems are fundamentally different from English. For example, some languages have only 6 vowels (a, e, i, o, u, y) – English has at least 14, not counting diphthongs. Some have a ‘flapped’ ‘r’ sound, while English has an approximant ‘r’ that doesn’t touch the roof of the mouth. Many languages have consistent stress on the penultimate syllable, whereas in English, stress is unpredictable and can change the meaning of words (REcord vs reCORD, PREsent vs preSENT).

All of this makes Listen and Repeat not just a pronunciation test, but also a test of auditory processing – you must hear sounds that your brain might automatically “filter out” and reproduce them in a fraction of a second.

How to prepare? The shadowing technique – listen to English podcasts, news, or TED talks and repeat after the speaker with a minimal delay (0.5–1 second). You don’t read the text; you listen and repeat. Practice 15 minutes a day for at least 6 weeks before the exam. On prepclass.io, you’ll find shadowing exercises specifically designed for the new TOEFL Speaking format, with instant AI feedback on your pronunciation.

Take an Interview: How to Speak Without Preparation

The second task of the new TOEFL Speaking section is Take an Interview, and the name says it all. You’re given a topic, followed by 4 questions on that topic. For each question, you have 45 seconds to respond. There’s no preparation time; the question appears on screen, you hear it through your headphones, and you speak immediately. Four questions, four answers, then it’s over.

The topics cover everyday life and student experiences: preferences for study styles, favorite campus spots, group work experiences, daily routines, travel, hobbies, future plans. These aren’t academic questions requiring specialized knowledge; they’re designed to check if you can spontaneously formulate coherent, elaborate answers in natural English.

Here’s what has changed compared to the old format: in the previous TOEFL Speaking version, you had 15–30 seconds to prepare your answer and could take notes. You could read a text passage, listen to a lecture, and create a response combining both sources. Now, none of that. There are no passages to read, no lectures to listen to, no notes. There’s just a question and 45 seconds.

The Structure of a 45-Second Response

Lack of preparation time doesn’t mean you can’t have a framework in your head. When you hear a question, your brain needs a schema, a ready-made structure that you’ll fill with ad-hoc content. This isn’t a template in the old sense (memorized phrases), but rather an answer architecture that works with any question.

Imagine the question: “Do you prefer studying alone or with a group? Why?”

Your answer should look like this: you open with a direct statement of your position (5 seconds), develop your first argument with an an example (15 seconds), add a second argument or contrast (15 seconds), and close with a brief summary or reflection (10 seconds). Total (45 seconds). Not too much, not too little).

Crucially: you don’t have to give the “correct” answer. The AI doesn’t evaluate whether you prefer group or solo study. It assesses whether you can express your opinion fluently, coherently, and with sufficient depth. A short “I prefer studying alone because it’s better” is not enough. An answer with a specific example (“Last semester, when I was preparing for my biology exam, I found that studying alone in the library helped me focus better than the study group sessions”) is what the AI is looking for.

Topic Categories: What to Expect

Based on official ETS materials and practice tests, Take an Interview topics revolve around several consistent areas. Student life is fundamental, with questions about choosing a major, time management, relationships with professors, and dormitory experiences. Personal preferences also appear, such as favorite ways to spend free time, dietary preferences, and communication style. The third category covers experiences and memories, with questions about travel, important decisions, and situations that taught you something.

There are no questions about politics, controversial social topics, or situations requiring specialized knowledge. ETS designs questions so that every test-taker, regardless of culture or experience, can answer them. As long as you’ve been a student, had hobbies, and made decisions, you have material for all the answers.

How to Practice Spontaneous Speaking?

The truth is, spontaneous speech in a foreign language is the most difficult skill to master, and the only one that can’t be learned from a textbook. You need hours of speaking practice in situations that don’t allow for preparation. On prepclass.io, you have a Take an Interview simulation mode: AI asks you questions in random order, and you speak into the microphone. After each answer, you receive a score and specific tips on what to improve.

But there are also methods you can use on your own, every day. The first is a timed monologue – set a timer for 45 seconds, randomly open a page with English conversation questions (e.g., “conversation starters” on Google), and speak. Don’t stop, even if you make a mistake; the TOEFL AI doesn’t penalize a single self-correction, but it does penalize long pauses.

The second method is recording and listening back. Record your answer on your phone, listen to it, and ask yourself: does it sound natural? Do I have long pauses? Do I use “ummm” and “ehhh” too often? Is my intonation flat? This is a brutal exercise, because your own voice always sounds strange, but it’s invaluable.

The third method is conversations with real people. If you have English-speaking friends, talk to them. If not, platforms like iTalki, Tandem, or HelloTalk give you access to native speakers for free or a symbolic fee. Goal: 20–30 minutes of conversation daily for 4–6 weeks before the TOEFL. This is the minimum for your brain to start automatically producing English sentences without mental translation from your native language.

You’ll find more preparation strategies for all sections in our complete guide to TOEFL 2026 and in the article on the Listening section, which is closely linked to Speaking, as Listen and Repeat is partly a listening test.

AI Scoring: How the Algorithm Evaluates Your Responses

One of the most controversial changes in TOEFL 2026 is the shift from human scoring to artificial intelligence scoring. In the old format, your recordings went to 3–6 human examiners who independently assessed each response. Now, an algorithm evaluates them, a machine learning model trained on hundreds of thousands of recordings previously scored by humans.

What does this mean in practice? The AI algorithm is more consistent than humans; it doesn’t have bad days, isn’t tired after 8 hours of listening, and doesn’t subconsciously give higher scores for an attractive voice. But it’s also more “literal”; if your “th” pronunciation is technically incorrect, the AI will catch it every time, whereas a human examiner might overlook it if the rest of your answer was excellent.

The algorithm scores your responses on a 1–6 scale, analyzing several dimensions. In Listen and Repeat, the key factors are: repetition accuracy (did you say the same words), pronunciation of individual sounds, intonation and word stress, pace, and fluency. In Take an Interview, the algorithm looks at: coherence and organization of the response, richness and precision of vocabulary, grammar (though minor errors don’t disqualify), pronunciation and intonation, and development of ideas (is the answer in-depth or superficial).

Important information: AI does not penalize accent itself. You can have a noticeable accent and still achieve a band 5 or 6, provided you are understandable. The algorithm distinguishes between “accented but understandable” and “accented and unintelligible.” If a native speaker would have to wonder what you said, that’s a problem. If a native speaker hears that you’re not American but understands every word, that’s not a problem.

Preparation Plan: From Pronunciation to Fluency

Preparing for the new TOEFL Speaking requires a fundamentally different approach than the old format. Forget about memorizing templates. Forget about practicing lecture notes. You need to work on two things: pronunciation and spontaneous speech. Here’s a realistic plan for 6–8 weeks before the exam.

Week 1–2: Diagnosis and Phonetic Awareness

Start by recording yourself. Read any English article aloud (2 minutes is enough). Listen to the recording and compare it with the original (e.g., an article read by a speech synthesizer). Identify your main pronunciation issues; for many non-native speakers, these will include: ‘th’ sounds, vowel length, ‘r’ sounds, and intonation. Write them down. This is your priority list.

Concurrently, start shadowing – listen to English podcasts (BBC 6 Minute English, TED Talks Daily, NPR Up First) and repeat after the speaker with a minimal delay (0.5–1 second). No pausing, no stopping the recording. You listen and speak simultaneously, trying to mimic the intonation, pace, and word stress.

Week 3–4: Working on Pronunciation and Fluency

Focus on your phonetic issues from the list. For “th” sounds – practice 5 minutes daily in front of a mirror: think, three, through, the, this, that, them, those. For vowels, minimal pairs: ship/sheep, bit/beat, pull/pool, cot/caught. For intonation, read sentences aloud, deliberately emphasizing content words and reducing function words.

Also, start spontaneous speaking exercises: daily, give 3 responses of 45 seconds each to random conversation questions. On prepclass.io, you have a question bank in the Take an Interview format: AI will assess your answer and indicate where you’re losing points. If you don’t have access to the platform, use a list of questions from the internet (search “TOEFL speaking practice questions 2026”) and record yourself on your phone.

Week 5–6: Full Section Simulations

Practice the entire Speaking section under exam conditions: 7 sentences to repeat + 4 questions to answer. On prepclass.io, you have full simulations with AI scoring, using technology identical to the real exam. Do a minimum of 3 full simulations per week. After each, analyze: what went well, where were the pauses, which sounds need improvement.

Week 7–8: Final Polishing

The last two weeks are for fine-tuning. Focus on areas that still need work. Practice shadowing with more challenging materials (academic lectures, debates). Do one full simulation daily. In the final week, reduce intensity so you don’t go into the exam fatigued. Review your response framework (Hook → Argument 1 → Argument 2 → Closing), and trust your preparation.

Remember: the Speaking section is just one part of the TOEFL. Simultaneously prepare for Reading, Listening, and Writing. All sections of the new TOEFL have undergone a revolution, and all require a different approach than before.

Exam Day Strategy

You now know how the new Speaking section works. You know what the AI assesses. You’ve practiced for weeks. Now, it’s exam day. Here are a few practical tips that could save your score.

Before you start Speaking, you have an advantage: the previous sections have warmed up your brain. The Speaking section is either last or second-to-last, depending on the test version. By this point, your brain has already been working in English for over an hour – you’ve read, listened, and written. This is a natural “warm-up” – use it.

In Listen and Repeat: don’t panic if you don’t hear the entire sentence. Repeat what you did hear. The algorithm awards partial points for partial correctness. “The professor suggested that students… should consider enrolling in the introductory seminar” is better than silence, even if you missed the middle of the sentence. AI rewards effort, it doesn’t penalize for imperfect memory.

In Take an Interview: start speaking immediately. Even if you don’t yet know what you’ll say next, begin with a statement of your position. “That’s a great question. I would say that…” – these 3 seconds give your brain time to formulate a response. AI doesn’t penalize for “filler” at the beginning; it penalizes long silence. Speak, even if you have to improvise.

Control your pace. A common mistake: you start too fast, blurt everything out in 20 seconds, and then have 25 seconds of silence. Or you start too slowly and get cut off mid-thought. The framework (5s + 15s + 15s + 10s) helps, but you need to practice it enough to intuitively feel the 45 seconds, without glancing at a timer.

If you also want to practice other standardized exams, such as the SAT, which many universities require alongside TOEFL, check out the adaptive tests on okiro.io.

A good strategy is also to properly compare TOEFL with other exams. If you’re hesitating between TOEFL and IELTS, check out our comparative guide TOEFL vs IELTS. And if you want to know what TOEFL score you need for specific European universities, read the article on TOEFL scores for studies in Europe.

Summary: New TOEFL Speaking Requires a New Approach

The Speaking section in TOEFL 2026 is the biggest change in the exam’s history. Four tasks replaced by two. Human examiners replaced by AI. Preparation time for responses, zeroed out. Listen and Repeat directly tests your pronunciation. Take an Interview tests your ability for spontaneous communication. There’s no longer room for memorized templates, pre-learned phrases, or “tricks” from a prep course.

For many international candidates, this is both a challenge and an opportunity. A challenge because pronunciation, intonation, and fluency are areas where non-native accents can be noticeable, and the AI algorithm will detect them. An opportunity because the section is shorter (8 minutes instead of 17), meaning less fatigue and less stress. And with proper preparation (6–8 weeks of systematic work on shadowing, spontaneous speaking, and phonetic awareness), a band 5 or 6 is fully achievable.

Next Steps

1. Take a diagnostic recording: read an English text, listen, identify problems
2. Start shadowing today: 15 minutes daily with BBC, TED Talks, or podcasts
3. Register on prepclass.io: Speaking simulations with AI scoring, Take an Interview question bank, pronunciation exercises with feedback
4. Read the other TOEFL 2026 guides: complete guide, Reading, Listening, Writing
5. Check university requirements: what TOEFL score for studies in Europe, TOEFL vs IELTS, which to choose
6. Practice daily: regularity beats intensity. 45 minutes a day for 6 weeks > 4 hours on a weekend

The new TOEFL Speaking tests how you truly speak English, not how well you can memorize templates. And that means the only path to success is actual speaking. Every day. Aloud. Into a microphone. Start now, because on the exam, no one will give you time to prepare.