The 2026 TOEFL iBT format, section by section
On 21 January 2026 ETS replaced the TOEFL iBT with an updated test. The new test is 120 items in under ninety minutes, two of its four sections are adaptive, and it reports on a 1–6 band scale. This guide lays the test out from ETS's own blueprint and from the behaviour of the official practice software — what is in each section, how long each part runs, what the screen lets you do, and how it is scored.
What changed on 21 January 2026
The left-hand column is ETS's own description of the test before 21 January 2026, from its archived test-content page; the right is the current page and the blueprint. Most preparation material still in circulation describes the left-hand column. The full before-and-after, including ETS's score comparison table →
| Until January 2026 | From 21 January 2026 | |
|---|---|---|
| Score | 0–120, four sections of 0–30 | 1–6 in half-band steps, per section and overall, aligned to the CEFR |
| Length | "Just under 2 hours" (ETS); plan for 2.5 hours with check-in | Base times of 30, 29, 23 and 8 minutes; ETS says to allow approximately two hours excluding directions; the blueprint gives 1 h 23 min – 1 h 29 min of test time |
| Section order | Reading, Listening, Speaking, Writing | Reading, Listening, Writing, Speaking |
| Reading | "Read passages and respond to questions" — 20 questions, 35 minutes (ETS) | 50 items across three task types; two-stage adaptive |
| Listening | "Answer questions about brief lectures or classroom discussions" — 28 questions, 36 minutes (ETS) | 47 items across four task types, from single exchanges to academic talks; two-stage adaptive |
| Speaking | "Talk about a familiar topic and discuss material you read and heard" — 4 tasks, 16 minutes (ETS) | Eleven items: seven sentences to repeat and a four-question interview; no task built on material read or heard |
| Writing | "Read a passage, listen to a recording, type your response" and "state and support an opinion in an online classroom discussion" — 2 tasks, 29 minutes (ETS) | Ten Build a Sentence items, an email, and the academic-discussion task |
| Marking | Not restated here: ETS's archived page does not describe it | 108 items machine-scored; the 12 written and spoken responses AI-scored (blueprint) |
The blueprint: four sections, 120 items, 145 raw points
ETS's published specification, one row per section. Item counts include a small number of unscored items in Reading and Listening that ETS uses to trial future questions; you are not told which they are.
| Section | Items | Task types | Marked by | Estimated time |
|---|---|---|---|---|
| Reading | 50 35 raw points | Complete the Words · Read in Daily Life · Read an Academic Passage | Machine | Router 18–21 min, then a 9 min module |
| Listening | 47 35 raw points | Listen and Choose a Response · Listen to a Conversation · Listen to an Announcement · Listen to an Academic Talk | Machine | Router 18 min, then 7 min (lower) or 11 min (upper) |
| Writing | 12 20 raw points | Build a Sentence ×10 · Write an Email · Write for an Academic Discussion | 10 machine, 2 AI-scored | 23 min |
| Speaking | 11 55 raw points | Listen and Repeat ×7 · Take an Interview ×4 | AI-scored | 8 min |
| Total | 120 145 raw points | Ten task types | 108 machine · 12 AI | 1 h 23 min – 1 h 29 min |
Source: ETS, TOEFL iBT Test: 2026 Update — Test Blueprint and Specifications. Timings are ETS's estimates and exclude the direction screens between sections.
Section 1 · about 27–30 minutes
Reading
Fifty items, three task types, two stages. The first stage is the router, which ETS times at 18 to 21 minutes; the second is a 9-minute module whose difficulty depends on how you did in the first. The task mix is the same in both.
- Complete the Words — 30 items, B1 to C1+
- A short academic text with words partly removed: the first letters stand and the rest is a row of blanks, one per missing letter. You type the letters. Ten blanks share a screen, the cursor moves on as each blank fills, and a blank's length tells you how many letters are missing. ETS classes this as reading for form as well as meaning: it tests whether you know the word the sentence needs, not whether you can guess a synonym.
- Read in Daily Life — 5 to 15 items, A1 to C1
- Short non-academic texts of the kind you meet outside a classroom — a notice, a message, a social-media post, a schedule — drawn as the real thing rather than as prose, with two or three questions each. The easiest texts are 15 to 50 words; the questions ask what the text is for, what it tells you, and what it implies.
- Read an Academic Passage — 5 to 15 items, B1 to C2
- A passage of up to about 200 words in the academic register, with multiple-choice questions on main ideas, details, inference, the relationship between ideas and the structure of the argument. A vocabulary question highlights its target word in the passage. The text stays on screen throughout.
What the screen lets you do
One question per screen, stimulus on the left, question and plain options on the right. Back and Next are both present, you may skip a question and return to it, and you can move freely inside a module. You cannot return to the router once the second module begins. The clock counts the module, not the question.
The full Reading guide: Complete the Words, and how to pace a module you can move around in →
Section 2 · about 25–29 minutes
Listening
Forty-seven items, four task types, two stages: an 18-minute router, then a second module of 7 minutes (lower) or 11 minutes (upper). Recordings run from a single utterance of a few syllables to a monologue of about 250 words, in a deliberate mix of U.S./Canadian, British and Australian voices, with a picture of the speaker on screen. Every recording plays once.
- Listen and Choose a Response — 15 to 19 items, A1 to B2
- One speaker says one thing; you pick the reply that fits. The recording is short — ETS caps it at six stressed syllables — so there is nothing to take notes on and no second hearing. It tests whether you caught the function of what was said: a request, an offer, a complaint, a question.
- Listen to a Conversation — 10 items, A2 to C1
- Two speakers, a short exchange on a campus or everyday matter, then questions on what was said, what was meant, and what happens next.
- Listen to an Announcement — 6 to 10 items, A2 to C1
- One speaker addressing a group — a change of room, a deadline, a service notice — and two questions on the essentials: what is changing, for whom, and what you are being asked to do.
- Listen to an Academic Talk — 8 to 16 items, A2 to C2
- The longest recording, a lecture-style monologue, followed by a set of questions on the main idea, supporting points, the speaker's attitude and, at the top of the range, idiomatic or less common vocabulary.
What the screen lets you do — and does not
This is the section that catches people. While a recording plays you see the speakers and nothing else: no question, no options, no Next. The questions follow, one per screen, each on its own clock. There is no Back, and the software refuses to advance while a question is unanswered. Answer, then move; there is no other order.
The full Listening guide: every task type, and where each goes wrong →
Section 3 · 23 minutes
Writing
Twelve items in three task types, not adaptive, and — for the first time — mostly machine-marked. Ten items are worth one point each and are scored by rule; the two written responses are worth five points each and are AI-scored. The clock reads hours, minutes and seconds, and each task is timed on its own: leave a task with time to spare and the software asks you to confirm, because you cannot come back.
- Build a Sentence — 10 items, A1 to C2
- A short exchange between two people. The reply is a sentence with its first and last words fixed and the middle missing; below it is a bank of words. You drag words into the slots to make a correct, appropriate sentence. In the official practice test the bank held one word more than there were slots, so one word was a distractor. Clicking a word does nothing; dragging is the only interaction.
- Write an Email — 1 item, B1 to C2, 7 minutes
- A scenario, an instruction, and a bulleted list of the things the email must do. You write into an editor with To and Subject lines already set, a live word count, and cut, paste, undo and redo. The prompt asks for as much as you can write, in complete sentences. It is scored 0 to 5 on whether it does what was asked, in a register that fits, with language that is accurate and varied.
- Write for an Academic Discussion — 1 item, B1 to C2, 10 minutes
- The task that survived from the old test. A professor's post sets a question; two students have already replied; you add your own contribution of at least 100 words, taking a position and supporting it. Same editor, same 0 to 5 scale.
The full Writing guide: why half the marks are sentences you assemble, and how each task is read →
Section 4 · 8 minutes
Speaking
Eleven items, two task types, eight minutes, all AI-scored out of five. There are no integrated tasks, no preparation time and no buttons: the microphone opens and closes itself on a visible countdown. A microphone check runs before the test starts, so a bad headset is found before it can cost you anything.
- Listen and Repeat — 7 items, A1 to C2
- A scenario is set — you are helping someone find their way, say — over a schematic illustration. You hear a sentence once and repeat it; each sentence relates to a different part of the picture, which is highlighted as you go. The sentences lengthen and grow more complex across the seven, from one clause to several. Scoring is on accuracy and intelligibility: the words, and the stress, rhythm and intonation that carry their meaning.
- Take an Interview — 4 items, A1 to C2
- A scenario screen, then an interviewer on video asks four questions aloud. The question is never shown in print, so you answer what you heard. Each answer has about three-quarters of a minute. The questions move from the factual to opinion, explanation and prediction, and the response is scored on how clearly and coherently you elaborate, with what grammar and vocabulary, and how intelligibly.
How it is scored
Two mechanisms. The 108 selected-response items — everything in Reading and Listening, and Build a Sentence — are scored by rule: right or wrong, one point each. The two written responses and the eleven spoken ones are scored by ETS's automated systems out of five each, on fluency, coherence, grammar and, for speech, intelligibility. That gives 145 raw points across the test.
What you receive is not raw points. Each section is reported as a band from 1 to 6 in half steps, and the overall score is the mean of the four, rounded to the nearest half band. ETS aligns the bands to the CEFR levels A1 through C2. The conversion from raw points to a band is done by equating against a calibrated item pool, and because Reading and Listening are adaptive, the same raw score means a different band depending on which second module you were routed to. ETS does not publish that conversion. Our practice reports estimate it and say so on the report; an official score report is the only place the real figure exists. For two years after January 2026 that report also carries a comparable overall score on the old 0–120 scale, and ETS publishes the comparison table behind it.
| Band | CEFR alignment |
|---|---|
| 6 | C2 |
| 5 – 5.5 | C1 |
| 4 – 4.5 | B2 |
| 3 – 3.5 | B1 |
| 2 – 2.5 | A2 |
| 1 – 1.5 | A1 |
The band-to-CEFR alignment is ETS's. Institutions set their own required band; check the figure your programme publishes rather than converting an old 0–120 target.
The full score guide: how 145 raw points become a band, and what an old 100 means now →
What the format means for how you prepare
- Pacing is the test.Ninety minutes for 120 items leaves under a minute an item on average, and Listening's per-question clock cannot be borrowed against. Sit at least one complete test under the real clocks before test day; a section practised on its own hides how the pacing compounds.
- Listening has to be answered in order, first time. Practising with replays or with the questions visible during the audio trains a habit the test will not allow.
- The productive tasks are short and scored on precision. A seven-minute email and a 45-second interview answer reward doing the asked-for thing accurately over doing more. Build a Sentence rewards grammar you can produce fast, not grammar you can explain.
- Old-format material trains the wrong reflexes. Integrated tasks, note-taking through lectures and the two-passage Reading are gone; time spent on them is time spent on a test you will not sit.
Questions about the format
How long is the 2026 TOEFL iBT?
ETS's blueprint gives a minimum test time of 1 hour 23 minutes and a maximum of 1 hour 29 minutes; its test-content page lists approximate base times of 30, 29, 23 and 8 minutes and says to allow approximately two hours to complete the test, excluding directions, adding that time and item counts vary as the test adapts. The spread is the adaptive branch: a Listening candidate routed to the upper module gets an 11-minute second module, the lower module is 7, and Reading's router is stated as 18 to 21 minutes.
Which sections are adaptive, and how does that work?
Reading and Listening. Everyone sits the same first module, called the router. How you do on it decides whether the second module leans easier or harder. Writing and Speaking are linear: everyone on a form gets the same tasks in the same order. Because the two branches differ in difficulty, the same raw score after the upper module is worth more than after the lower one — which is why a raw score alone does not tell you your band.
Can I go back to a previous question?
In Reading, yes, within the module you are in: there are Back and Next controls and you may leave a question unanswered and return to it. Once a module ends you cannot return to it. In Listening there is no Back at all, each recording plays once, and the software will not let you move on with a question unanswered. In Writing and Speaking each task is timed on its own and moving on is final.
What is the new score scale?
Each section is reported as a band from 1 to 6 in half steps, and the overall score is the mean of the four sections rounded to the nearest half band. ETS aligns the bands to the CEFR — broadly, 6 to C2, 5 and 5.5 to C1, 4 and 4.5 to B2, 3 and 3.5 to B1. ETS does not publish the conversion from raw points to a band; it equates scores against a calibrated item pool. Any raw-to-band ladder you see outside an official score report, including ours, is an estimate.
Are the old 0–120 scores and the old preparation books still useful?
Scores from tests taken before 21 January 2026 remain valid for two years from the test date and institutions continue to accept them while they republish requirements on the new scale. Preparation written for the old test is another matter: the integrated speaking and writing tasks, the two-passage Reading and the lecture-heavy Listening no longer exist, so material built around them trains a test you will not sit. Vocabulary and grammar work carries over; task practice does not.
Which accents are used in Listening?
ETS states that the recordings use a balanced mix of U.S./Canadian, British and Australian English, delivered by AI-generated voices and sometimes human voice actors, with a visual of the speaker on screen. Each recording plays once.
Is there a break?
No. ETS's test-content page states that there are no scheduled breaks during the TOEFL iBT test and advises planning accordingly.
Can I take notes?
Yes. ETS's test-content page: at a test centre, scratch paper and pencils are provided; for the Home Edition, use a small whiteboard or a blank sheet of paper in a transparent sheet protector with an erasable marker. ETS's Listening page adds that you can take notes while listening. Writing-assistance software is not permitted.
Can I still take it at home?
Yes. ETS offers the test two ways: on a computer at an authorised test centre, or the TOEFL iBT Home Edition on a computer at home, monitored by a live human proctor. The Home Edition's equipment and room rules, and the fee and deadlines for either route, are on the registration page.
Sources: ETS, TOEFL iBT Test: 2026 Update — Test Blueprint and Specifications (item counts, raw points, CEFR targets and estimated timings), and the behaviour of the official TOEFL iBT practice interface as observed (navigation, clocks and on-screen controls). Checked 5 September 2026. ETS notes that the specification may see minor revisions; the blueprint is the authority. ExamStep is an independent practice resource and is not affiliated with ETS. TOEFL® and TOEFL iBT® are registered trademarks of Educational Testing Service.