TOEFL 2026 Speaking Section:
The Complete Guide

What it looks like: You see an image related to a campus or daily life scenario — a library tour, a laboratory safety briefing, directions to a campus office, or a step-by-step process. You then hear seven sentences, one at a time, that describe the scenario illustrated in the image. After each sentence, a beep sounds and you must repeat the sentence exactly as you heard it. The sentences get progressively longer and more complex across the seven items.

The task tests pronunciation, intonation, rhythm, phonological memory, and accuracy. It does not test comprehension or knowledge — the image provides context that helps you understand and remember the sentence, but the only thing being evaluated is how accurately and clearly you can reproduce what you heard.

Official ETS scoring rubric: Listen and Repeat

Source: ETS official TOEFL 2026 scoring documentation at ets.org/toefl/test-takers/ibt/scores.html

Strategy for Listen and Repeat

• 1 Use the image as a memory scaffold The image is there for a reason. Before the first sentence plays, look at it carefully and understand the scenario — a library tour, a lab procedure, a campus directions sequence. When you hear each sentence, the image gives your brain context that dramatically improves your ability to retain and reproduce the words accurately. Students who ignore the image and focus only on the audio are missing a significant memory aid.
• 2 Practice shadowing every day Shadowing means listening to a sentence and repeating it in real time — overlapping with the speaker slightly rather than waiting until the sentence is completely finished. This is the most direct training for the phonological memory and real-time reproduction skills that Listen and Repeat tests. Use authentic English speech — podcasts, news broadcasts, academic lectures — and practice shadowing individual sentences for 10 to 15 minutes daily.
• 3 Focus on problem sounds specifically Identify the English sounds that are genuinely difficult for your phonological system — for Spanish speakers this often includes the difference between v and b, the th sounds, and specific vowel distinctions. For Italian speakers it often includes the h sound and certain consonant clusters. Targeted pronunciation work on your specific problem sounds produces faster improvement than general pronunciation study.
• 4 If you lose a word, self-correct immediately The ETS rubric explicitly recognizes self-correction as a positive signal. If you realize you said the wrong word, correct yourself in real time and continue. A corrected error scores better than an uncorrected one. Never stop speaking to think — keep moving through the sentence.

What it looks like: You see a short looping video of an interviewer who introduces the topic. You are told you have volunteered for a research study on a familiar everyday topic — daily routines, entertainment, city living, travel, technology, education, or similar accessible subjects. The interviewer then asks you four questions that progressively increase in complexity and abstractness. You hear each question once and must begin speaking immediately after the beep. You have 45 seconds to respond to each question.

The four questions within each interview follow a consistent progression. The first question is typically concrete and personal — what do you currently do, or what is your current situation regarding this topic. The second asks for your preferences or habits with a reason. The third asks for your evaluation or opinion — advantages, disadvantages, comparisons. The fourth is typically the most abstract — how might this change in the future, or what are the broader implications.

Official ETS scoring rubric: Take an Interview

The Take an Interview task is scored using four constructs. Each response is evaluated holistically across all four, resulting in a single 0 to 5 score per response. According to ETS official scoring guidance:

The 45-second structure that scores well

The most effective and consistent structure for Take an Interview responses uses three phases that correspond roughly to time segments within the 45 seconds:

Seconds 0 to 15 — State your main idea clearly. Answer the question directly in the first sentence. Do not introduce yourself, restate the question, or add filler. The scorer is listening for your answer from the very first word. A response that opens with "Well, that is a really interesting question" has already lost approximately three seconds and signaled scripted delivery.

Seconds 15 to 38 — Develop with one specific example or reason. Elaborate on your main idea with one concrete supporting point. A specific example from personal experience, a real observation, or a brief explanation of why your answer is true. Specificity is rewarded — a concrete detail ("I take the metro to university every morning, which takes about 35 minutes") scores better than a vague generalization ("I use public transport because it is convenient").

Seconds 38 to 45 — Close with a brief concluding point or contrast. Round off the response naturally. This does not need to be elaborate — a single sentence that acknowledges a limitation, adds a contrasting idea, or brings the response to a natural close is enough. What matters is that the response does not simply stop mid-thought when the timer ends.

Sample interview question and model answer

The following topic and question are representative of the style and difficulty level of real TOEFL 2026 Interview tasks. The model answer demonstrates the 0 to 15 / 15 to 38 / 38 to 45 structure at approximately band 5.0 level.

Three more sample interview questions for practice

Strategy for Take an Interview

• 1 Answer the question in your first sentence, always The most common mistake on Take an Interview is opening with a statement that does not answer the question. "That is an interesting point" or "There are many views on this topic" wastes time and signals scripted delivery. Your first sentence must be your answer. Train this habit from the very first day of practice — it takes two to three weeks to become automatic but produces immediate score improvement once it does.
• 1 Fill the 45 seconds — this is a measurable target At 150 words per minute, a 45-second response contains approximately 110 to 115 words. Count your words in practice recordings. If you are consistently under 90 words, you are not developing your answers enough. If you are consistently over 120 words, you may be rushing. The target range is not arbitrary — it reflects the fluency rate the scoring engine is calibrated to reward.
• 3 Use specific details rather than generalizations "People in big cities tend to be stressed" scores lower than "I noticed when I was in London last summer that people on the tube rarely made eye contact or spoke to each other, which felt quite different from Barcelona." Specific details — places, numbers, personal observations — demonstrate the authentic language use that both the AI scoring engine and human raters are trained to reward.
• 4 Record yourself every day and listen back critically Most students who practice without recording cannot hear their own fluency gaps, pronunciation issues, or organizational weaknesses. Your ear compensates for what your mouth does imperfectly. Recording and listening back is genuinely uncomfortable but produces faster improvement than any other single practice habit. Focus on three things each time: did I answer the question directly, did I fill the time, and was my speech continuous without long pauses.
• 5 Do not rehearse topic-specific content The scoring engine is specifically designed to detect memorized content. Students who prepare ten prepared answers on different topics and then adapt them to whatever question comes up score lower than students who respond genuinely in the moment. The authenticity of spontaneous speech is a scorable signal. Prepare the structure and the habits — not the content.