How to Study for the TOEFL: The Test Changed in 2026

To study for the TOEFL, first check which test your materials describe. ETS redesigned the TOEFL iBT on 21 January 2026, so scores now run 1 to 6, Reading and Listening adapt as you answer, and you spend about ninety minutes answering questions inside the two hours ETS asks you to allow. Then build your plan around vocabulary and sentence form, because one task type is half the Writing score outright and another is the largest block of items in Reading.

That opening is worth more than it looks, because almost every TOEFL guide, review book and AI answer in circulation still describes the previous test. If your preparation is organised around a 0 to 120 total, four sections marked out of 30, or the integrated tasks that asked you to combine a passage or a recording with your own spoken or written response, you are studying for a test that is no longer administered. This page is written from the live ETS pages as they stand in September 2026, with the per-task breakdowns taken from the 2026 specification document and flagged wherever they are.

Test length is the trap inside that list, because it will not announce itself. The previous test was not the three-hour exam of TOEFL folklore. Seven weeks before the change, in December 2025, ETS was already telling test takers that “The total test takes just under 2 hours to complete, but you should plan for 2.5 hours, allowing 30 minutes for check in”, and its sections ran Reading 35 minutes, Listening 36, Speaking 16 and Writing 29. The live page no longer carries that wording, so the quotation above is from the archived ETS test content page of 3 December 2025. That is about 116 minutes of content, against roughly 90 today. So a 2024 review book claiming a two-hour test looks current and is nothing of the kind. Audit your materials on the scale and the task names instead, since those two changed unmistakably.

What changed on 21 January 2026?

The headline is the scale. The ETS page on understanding TOEFL scores dates the change precisely, saying that as of January 21, 2026, TOEFL iBT score reports feature a new 1 to 6 scale, in half point increments, with an alignment to the Common European Framework of Reference for Languages. You now receive four section scores and one overall score on that scale.

The content moved at the same time, and the section pages are blunt about the direction. The Reading section page says that “Instead of long textbook passages, the Reading section uses focused texts and interactive tasks”, and the Listening section page mirrors it with “Instead of long listening sets, the Listening section uses short, focused audio clips”. Alongside that, Reading and Listening became adaptive, and the whole sitting shrank to roughly an hour and a half of questions.

The consequence a returning test taker most often gets wrong is what happened to the integrated tasks. Those were the four that defined old TOEFL preparation, where you combined source material with a spoken or written answer of your own, and they do not appear in the current published task lists. The Speaking section page names Listen and Repeat and Take an Interview. Better evidence than any of the section pages sits in the specification document, which carries an Allowable Item Types row for each section, a closed list by construction. The Listening row reads Listen and Choose a Response, Listen to a Conversation, Listen to an Announcement, Listen to an Academic Talk, and no section row anywhere contains an integrated task. The per-task item counts also sum to the section totals ETS publishes, which leaves the old templates nothing to attach to.

How is the TOEFL scored now?

Each of the four sections earns a score on the 1 to 6 scale in half point steps, and the overall score is, in the published wording, “the average of all four sections, rounded to the nearest half band”. ETS gives two worked examples across its pages, and reading them together is more instructive than either alone. On the scores page, an average of 5.25 is awarded an overall 5.5. On the test content page, an average of 5.125 is awarded an overall 5. The first rounds up to the nearest half band and the second rounds down to the nearest whole one, which together show the grain of the scale: your four section scores land on a half point grid, and the average of them is pushed back onto that same grid.

If that mechanism feels familiar, it is because it is now essentially the IELTS one. Both exams average four equal section scores and round to the nearest half band, and the arithmetic that follows has a consequence people routinely get backwards. Our guide to studying for the IELTS works the derivation through in full, so this page will point at it instead. The short statement is that averaging four equal sections plays no favourites. A half point won in Speaking lifts your overall score by precisely what a half point won in Reading lifts it. So when a guide tells you to prioritise your weakest skill and offers the scoring maths as its reason, the maths does not actually say that.

The scale change also reaches into how you should read a university admissions page this year. Start with the transition. For a two-year period after January 2026 you also receive a comparable overall score on the 0 to 120 scale, representing the midpoint of the corresponding total range, and ETS publishes a comparison table on the scores page mapping each 1 to 6 band onto the older section and total ranges. The rest of it is that the design deliberately spares you the conversion. In the words of the scores page, your official report includes “both the updated 1 to 6 scores and a comparable 0 to 120 total score estimate, so you can meet either format without converting anything”. Where an institution still lists a 0 to 120 requirement, you submit as usual.

Worth knowing alongside that: ETS states plainly that “There are no passing or failing scores set by the TOEFL Program or ETS”, and that each institution sets its own requirement. MyBest Scores also survived the change and now crosses the scales, so on a report from the new scale the MyBest section converts older 0 to 120 results onto the 1 to 6 scale, while a report on the prior scale includes prior-scale results only.

Where the items actually sit

Here is the part worth planning your weeks around, and seeing it takes reading past the summary table. The test content page publishes that table: Reading is 50 items in about 30 minutes, Listening 47 items in about 29 minutes, Writing 12 items in about 23 minutes, and Speaking 11 items in about 8 minutes.

Those item counts invite a wrong assumption. Twelve Writing items sounds like twelve pieces of writing. It is nothing of the sort, and the TOEFL iBT 2026 test specifications document breaks each section down by task type. Read it with one caveat in mind, which ETS supplies itself: the document says its specifications are “subject to minor revisions as part of the ongoing development and validation process until the official launch of the assessment”. It is a pre-launch specification and the launch has happened, so treat the live pages as authoritative where the two differ. The section totals in the specification match the live table exactly, which is reassuring, and the internal breakdown goes like this.

• Reading, 50 items, 35 raw points. Thirty of the items are Complete the Words, a task the Reading page describes as completing partially missing words in a short text, measuring “your vocabulary knowledge and your ability to understand meaning from context”. The rest split between Read in Daily Life and Read an Academic Passage. Mind the gap between 50 and 35: the specification says “Max points per item is 1 for scored items in the Reading section”, so 15 of those 50 items do not count toward your score at all.
• Writing, 12 items, 20 raw points. Ten of the items are Build a Sentence, which the Writing section page describes as arranging words or phrases to form a complete and grammatical sentence or question, and the specification puts each one at a maximum of 1 point. Exactly two are extended writing, one Write an Email and one Write for an Academic Discussion, and each of those carries a maximum of 5. So Build a Sentence is 10 of the 12 items and precisely half of the Writing points.
• Speaking, 11 items, 55 raw points. Seven are Listen and Repeat, where you hear short sentences and repeat them exactly as you hear them. Four are Take an Interview, a simulated interview about academic or campus situations. Each item carries a maximum of 5 points.
• Listening, 47 items, 35 raw points, spread across Listen and Choose a Response, Listen to a Conversation, Listen to an Announcement, and Listen to an Academic Talk. Scored items are worth 1 point each here too, so 12 of the 47 do not count.

Before acting on any of that, note where the arithmetic misleads, because most of the obvious wrong turns flatter the conclusion.

The non-scored items come first, and they are not a rounding error. Blueprint note 1 says “Items include non-scored items in the reading and listening sections”, whose purpose is typically to gather data on item performance for future use. That is 27 of the 120 items you will answer, and ETS does not identify them. Since you cannot pick them out, answer everything as though it counts.

The other caution is that raw points do not compare across sections. Speaking carries 55 and Reading 35, which does not make Speaking the more valuable section, because each section maps to its own 1 to 6 band and the four bands are then averaged equally. A points total means something only inside the section it belongs to.

Inside those limits, one finding survives cleanly and the other is bounded. Build a Sentence is half the Writing score outright, and no essay template touches it. Complete the Words is the largest single block of Reading items, though ETS does not publish which task types the 15 non-scored Reading items are drawn from, so its real share of the Reading band sits somewhere between 43 and 86 percent and the published specification cannot narrow it further. Both point the same way. Vocabulary and sentence form carry more of this test than any preparation built for long passages and integrated essays would suggest, and they are also the part that responds fastest to drilling, since word knowledge and sentence patterns can be practised item by item in a way that fluency and argument quality cannot.

Building that vocabulary is a separate skill from knowing you need it, and two existing guides go deep on the how, so this page will point instead of repeat. Memorising vocabulary takes on the encoding problem, meaning how a word gets from a list into recall. Studying a foreign language takes on the slower business of lifting your general level, which is ultimately what a CEFR-aligned score is measuring. For the Listen to an Academic Talk items, the habits in studying from recorded lectures carry over almost unchanged, since that task is a short lecture and ETS confirms you may take notes while it plays.

What does a two-stage adaptive section mean for your preparation?

The specification document states that “both the Reading and Listening sections are designed as two-stage adaptive tests”, and that timing estimates should include the router plus the lower or upper module. In plain terms, the opening stretch of each of those two sections works out roughly where you are, and your answers there route you into either a lower or an upper module for the rest of it. Writing and Speaking work differently: the same document says those sections “are linear, where all test takers of a specific form receive the same set of tasks”.

What follows from that cuts against habits carried over from a linear test. The early items in Reading and Listening do more than earn their own marks, since they also decide which module you see, so warming up slowly across the first few questions costs more here than it did on a fixed form. A practice test from a book, meanwhile, cannot reproduce any of this. A paper test has one fixed difficulty path, so it can rehearse the task types and the clock, and it cannot rehearse the routing. Expect the real sitting to feel harder or easier partway through than any practice paper did, and treat that shift as the mechanism working.

Because the difficulty moves, and because some of the items you answer are never scored, a tally off a practice paper means much less here than a raw score out of 40 means on IELTS. What still works is dissecting the answers you got wrong, which is the part of practice testing that survives any format change. Our guide to reviewing a practice test is built around exactly that habit.

Two scoring engines, and what each one can see

The specification document also splits every task type by how it is scored, and the split is cleaner than the old test ever offered. Machine-scored task types use rule-based algorithms on selected-response formats, judged on exact matches or predefined scoring logic. The constructed responses, meaning the speaking and writing you produce yourself, are described as “evaluated using natural language processing and machine learning to assess features like fluency, coherence, and grammar”. Both of those descriptions come from the pre-launch specification, which no live page restates, so hold them as the published design of the test. What the live scores page adds is that TOEFL tests are “only scored by a centralized scoring team, never at the test center”.

That last row is a genuine decision point, and it collides with a deadline most people meet without noticing it is one. Your test fee includes up to four free official score reports, and recipients can be added through your ETS account only until 10 p.m. on the day of your test, which is before any Speaking or Writing score exists. Separately, you cannot request a score review once you have asked for scores to be sent to any institution or agency. Put those two together and using the free reports forecloses the review entirely. The 30-day clock is generous; the binding decision lands on test day, when the free sends and the right to challenge a Speaking or Writing score turn out to be one choice, made before you know whether you want either.

The unofficial-scores row matters for a different reason. Seeing Reading and Listening at the end of the test, with Speaking and Writing still pending, is a direct consequence of the two engines. ETS notes these remain unofficial because all scores go through quality checks and it is possible, though rare, for one to change.

Practical rules from the ETS pages that a study plan should absorb

A handful of published rules change how you rehearse, and each is worth building into a practice session early, because discovering one on the day is expensive.

• Finishing every section is non-negotiable. ETS states that if you do not complete at least one Reading question, one Listening question, one Speaking task and one Writing task, your test will not be scored, and puts it starkly: skipping any section means no sections will be scored.
• There are no scheduled breaks. Ninety minutes of questions runs straight through, which is worth rehearsing at full length at least once.
• You can take notes. At a test centre, scratch paper and pencils are provided; at home, ETS specifies a small whiteboard or a blank sheet in a transparent sheet protector with an erasable marker.
• Writing assistance software is not permitted during the test, which is a rule worth respecting in practice too, for reasons the next section gets into.
• Accents vary by design. The Speaking and Listening pages both note you may hear native-speaker English accents from North America, the U.K., New Zealand, or Australia, so build listening practice from more than one of them.
• Where you sit it. The TOEFL iBT test runs on a computer at an authorized test center, and the TOEFL iBT Home Edition runs on a computer at home monitored by a live human proctor.

What a study app can carry on the TOEFL, and what it cannot

Start with the refusals, because on this exam they are substantial and one of them is a test rule. GeniusPal cannot hear you, so it does nothing at all for a Speaking section whose two task types are repeating sentences aloud and answering interview questions. It cannot score your writing the way the test does, since those two extended tasks are evaluated by natural language processing and machine learning against criteria ETS has not published as a public rubric. It cannot simulate a two-stage adaptive section, because the routing is the whole mechanism and no question generator has it. And it is a study tool for before the test, never during: writing assistance software is prohibited in the exam itself.

The actual function is narrow. Hand it a document with a real text layer and it writes questions from what it finds inside. Accepted formats are PDF, .pptx, .docx, .txt, .md, .markdown, and .csv, up to 10 MB, with no image upload and no audio. There is no OCR, so a photographed page of a coursebook or a scan of a handwritten vocabulary notebook gives it nothing to work from.

Inside that boundary, the 2026 redesign happens to have moved the test toward the one thing this kind of tool does well. Word knowledge and sentence form are a large and highly drillable part of this exam, and they are built from material that already lives in documents: word lists, collocations, academic phrase banks, the vocabulary from texts you have read. Those convert into questions cleanly. The comprehension items and the listening items are a weaker fit, and the two produced halves of the test are outside it entirely.

That fit comes with a real limit on a free account. A quiz on a free account is entirely multiple choice, which exercises recognition, and recognition is the weaker half of what Complete the Words and Build a Sentence are asking for. Both of those tasks want you to produce the right form, unprompted, in context. Producing a word unprompted is what the written short-answer mix and the recall mode do, and both of those sit on the paid plans. So on a free account, treat this as vocabulary triage, useful for finding out which words have not landed. Closing the distance between recognising a word and producing it costs either a paid plan or a pen and a blank page.

The numbers, so none of it surprises you. A free account gets 2 study-set generations for the whole life of the account, at 10 questions per set, and each set can be sat 2 times through. Quiz mode and the daily review are free on every plan, and on a free account a quiz is entirely multiple choice. Flashcards, written active recall, sets of up to 30 questions, and written short-answer items inside a quiz all need a paid plan. Student is $14.99 a month for 100 generations a month, and Genius is $59.99 a year, shown as $5.00 a month, with a generation allowance running up to a fair-use ceiling.

Putting it together

Check the date on every resource you own before you use it, because 21 January 2026 split the TOEFL into two tests that share a name, and the one your book describes may well be the retired one. Work out the overall score your institutions want on the 1 to 6 scale, remembering that your report carries the 0 to 120 comparable for the next couple of years and that no institution requires you to convert anything. Then spend the bulk of your hours on vocabulary in context and sentence construction, which carry more of this exam than an older study plan makes room for. Rehearse one full ninety-minute run with no break, on more than one English accent, and answer something in every section even if a question defeats you, since skipping one voids the whole test. Decide before test day whether you might want a Speaking or Writing score reviewed, because sending your four free reports signs that right away. And note that scores expire at two years, with no extension to ask for.