Reference
The TOEFL® iBT format from January 2026: a sourced reference
What is the TOEFL iBT format from January 2026?
From January 21, 2026, the TOEFL iBT specification lists 120 items across four sections worth 145 raw points, with listed test time of 1 hr 23 min to 1 hr 29 min. Reading and Listening are two-stage adaptive and include non-scored items; Writing and Speaking are linear. The specification states its values are subject to minor revision until launch.S1
ETS, TOEFL iBT Test: 2026 Update — Test Blueprint and Specifications Document. Read .
This page records the structure of the TOEFL iBT test as ETS documents it, task by task. Every statement below is either quoted from one of the three sources listed in section 9 or is a listed value read from one of them, and each carries the date that document was read. Where the sources are silent, this page says so rather than filling the gap.
1.Scope, method and limits
The format described here applies to tests taken on or after January 21, 2026. The three sources were read on , and the method used for each is recorded beside it in section 9 so the reading can be repeated rather than trusted.
Two limits apply to everything below. First, the specification states its own values are provisional:
“Please be advised that the test specifications outlined in this document are subject to minor revisions as part of the ongoing development and validation process until the official launch of the assessment.”
Second, the listed item counts for Reading and Listening are not counts of scored items:
“Items include non-scored items in the reading and listening sections. Their purpose is typically to gather data on item performance for future use in scored sections of the test.”
This page is not ETS material and is not endorsed by ETS. It quotes published documents under nominative reference and links each one so a reader can check it.
Reach120 is an independent practice tool. It is not affiliated with, endorsed by, or approved by ETS, and it does not provide official TOEFL® test scores.
TOEFL® is a registered trademark of ETS. This product is not endorsed or approved by ETS.
2.The four sections
Listed values and section claims from the specificationS1, read August 23, 2026.
| Section | Items | Raw points | Estimated time | Design |
|---|---|---|---|---|
| Reading | 50* | 35 | 18–21 min router, then a 9 min module | Two-stage adaptive |
| Listening | 47* | 35 | 18 min router, then a 7 min lower or 11 min upper module | Two-stage adaptive |
| Writing | 12 | 20 | 23 min | Linear |
| Speaking | 11 | 55 | 8 min | Linear |
| Total | 120* | 145 | 1 hr 23 min – 1 hr 29 min | — |
* Includes non-scored items, per the note quoted in section 1. Estimated times for Reading and Listening are a router plus one module; see section 7 on the second time figure ETS publishes.
What each section claims to measure
The specification’s own claim statements, verbatimS1.
- Reading
- “The test taker can read and understand academic and nonacademic written texts presented in a variety of formats.”
- Listening
- “The test taker can understand spoken English in academic and social contexts, including single exchanges, short conversations, announcements, and academic talks, in order to identify meaning, purpose, and appropriate responses.”
- Writing
- “The test taker can produce grammatically accurate and contextually appropriate written English across a range of academic and interpersonal situations, including sentence-level construction, multi-sentence responses, and extended academic discourse.”
- Speaking
- “The test taker can produce intelligible and coherent spoken English to effectively communicate in both brief and extended interactions across general and academic contexts.”
3.The task inventory
The section totals above are assembled from twelve task types. The specification’s per-section tables give each one an item count or range, a CEFR level target, a language-use context and a scoring route; the per-section scoring notes give the maximum points for one item of that type.
| Task type | Items | CEFR target | Language-use context | Scoring | Max pts/item |
|---|---|---|---|---|---|
| Reading | |||||
| Complete the Words | 30 | B1–C1+ | Academic language use | Machine Scored | 1 |
| Read in Daily Life | 5–15 | A1–C1 | Social Interpersonal language use | Machine Scored | 1 |
| Read an Academic Passage | 5–15 | B1–C2 | Academic language use | Machine Scored | 1 |
| Listening | |||||
| Listen and Choose a Response | 15–19 | A1–B2 | Social Interpersonal language use | Machine Scored | 1 |
| Listen to a Conversation | 10 | A2–C1 | Social Interpersonal language use | Machine Scored | 1 |
| Listen to an Announcement | 6–10 | A2–C1 | Academic Navigational language use | Machine Scored | 1 |
| Listen to an Academic Talk | 8–16 | A2–C2 | Academic language use | Machine Scored | 1 |
| Writing | |||||
| Build a Sentence | 10 | A1–C2 | Social Interpersonal language use | Machine Scored | 1 |
| Write an Email | 1 | B1–C2 | Academic Navigational language use | AI Scored | 5 |
| Write for an Academic Discussion | 1 | B1–C2 | Academic language use | AI Scored | 5 |
| Speaking | |||||
| Listen and Repeat | 7 | A1–C2 | Academic Navigational language use | AI Scored | 5 |
| Take an Interview | 4 | A1–C2 | Academic Navigational language use | AI Scored | 5 |
Source: specification per-section tables and scoring notesS1, read August 23, 2026. Five task types carry a listed range rather than a fixed count.
What the two scoring routes mean
“Machine Scored task types use rule-based algorithms to evaluate selected-response formats (e.g., multiple choice), based on exact matches or predefined scoring logic. AI Scored task types involve constructed responses (e.g., speaking and writing) and are evaluated using natural language processing and machine learning to assess features like fluency, coherence, and grammar.”
Reach120 reads
The split does not follow the section boundaries, which is the most common misreading of it. Ten of the twelve Writing items are machine-scored; the two AI-scored Writing tasks and the eleven AI-scored Speaking items are the whole of the constructed-response test. Practice feedback on a constructed response — including ours — is an estimate, and is never an official score.
Why two of the sections have no fixed length
“To measure language proficiency efficiently, both the Reading and Listening sections are designed as two-stage adaptive tests. The TOEFL Writing and Speaking sections are linear, where all test takers of a specific form receive the same set of tasks. Total timing estimates should include the router plus the lower or upper module.”
4.The Writing section, and what changed
The Writing section is where the restructure is largest and where published guidance is most often out of date. The left column below describes what preparation material commonly says; it is not attributed to ETS, and neither ETS source consulted here names the pre-2026 tasks or states that any current task replaced a particular earlier one.
Commonly described as
Preparation material widely describes a Writing section of two long constructed responses: an integrated task built on a reading passage plus a lecture, and an independent opinion essay.
What the source states
The specification lists the Writing section as 12 items across three task types — Build a Sentence (10 items), Write an Email (1), Write for an Academic Discussion (1) — with 20 maximum raw points and an estimated 23 minutes.S1(E-003)
Commonly described as
That material treats the whole Writing section as constructed response, scored by rubric.
What the source states
Ten of the twelve listed Writing items are Build a Sentence items, which the specification marks Machine Scored and worth a maximum of 1 point each; the two constructed responses are marked AI Scored and worth a maximum of 5 points each.S1(E-016)
Commonly described as
It reports a Writing score out of 30 and a total out of 120.
What the source states
For tests taken on or after January 21, 2026, score reports use a 1–6 scale in half-band increments, with the overall score the average of the four sections rounded to the nearest half band.S3(E-001)
What one Writing or Speaking item is worth
“Max points per Build a Sentence item is 1. Max points for each Write an Email and Write for an Academic Discussion item is 5.”
“Max points for each Listen and Repeat and Take an Interview item is 5.”
Reach120 reads
Ten Build a Sentence items at a maximum of 1 point, plus two items at a maximum of 5, reconcile to the 20 raw points the blueprint lists for Writing. That is a reconciliation of listed maxima, not a statement about how a raw total becomes a reported band — see section 6.
5.Administration and the length of a sitting
“TOEFL iBT test: on a computer at an authorized test center. TOEFL iBT Home Edition: on a computer at home, monitored by a live human proctor.”
“Test time does not include directions. Please allow approximately two hours to complete the test. As the test adapts, test time and items may vary.”
“If you don’t complete at least one Reading question, one Listening question, one Speaking task and one Writing task, your test will not be scored. In other words, skipping any section means no sections will be scored.”
“TOEFL iBT test items are developed through a collaborative effort of researchers, test developers, and psychometricians using a multi-step process. AI-assisted tools may support development, but all items undergo rigorous human review.”
6.How results are reported
Score reporting is summarised here only far enough to keep the format intelligible. The scale itself, including the band comparison table ETS publishes for tests taken before January 21, 2026, is covered on the score-scale page.
“The overall score will be the average of all four sections, rounded to the nearest half band (for example, if the average of your section scores is 5.25, you will receive an overall score of 5.5).”
“For a two-year transition period after January 2026, you will also receive a comparable overall score on the 0 – 120 scale, representing the midpoint of the corresponding total range.”
Reach120 reads
Raw points and reported bands are different quantities. The specification lists raw maxima per item; the score guidance describes a 1–6 band. Neither document publishes a conversion between the two, so any raw-to-band table found elsewhere is somebody’s model, and this page does not print one.
7.Where the sources disagree
Two ETS documents give different section times, and both are current. Recorded rather than resolved.
SpecificationS1
The specification prints Reading and Listening as a router plus a module — 18–21 min plus 9 min for Reading, 18 min plus 7 or 11 min for Listening — and gives total test time as 1 hr 23 min to 1 hr 29 min.
Test-content pageS2
The test-content page prints a single “Approx. Base Time” per section instead: Reading 30, Listening 29, Writing 23, Speaking 8 minutes.
Reach120 reads
The two are consistent if the content page’s figure is read as the longest route through an adaptive section — 21 + 9 for Reading, 18 + 11 for Listening — and the Writing and Speaking figures match outright. That reconciliation is ours. ETS states neither it nor which figure supersedes the other, so a page that needs the shorter route should cite the specification and a page that needs the sitting length should cite the content page.
8.Checked, and not found
Each row below records something a reader may reasonably expect the sources to state, and what was on those pages instead when they were read on . An absence is only useful if it names which documents were checked and when.
- A required or recommended word count for Write an Email or Write for an Academic Discussion.
- Neither the specification nor the test-content page states any word count for either task. The specification describes the writing stimuli and the expected register at length and never gives a length.S1, S2
- A conversion from the specification’s raw points to a reported 1–6 band.
- The specification lists raw points; the score page describes the 1–6 scale. Neither publishes a mapping between them, and the score page’s comparison table converts bands to the retired 0–30 and 0–120 ranges, not raw points to bands.S1, S3
- A statement from ETS that a current Writing task replaces, or corresponds to, a particular pre-2026 task.
- Neither source names the pre-2026 Writing tasks at all. Any sentence of the form “the Academic Discussion task replaced the independent essay” is somebody’s reading, including ours where we make it.S1, S2
- A fixed number of questions a given test taker will see.
- The specification prints ranges for five of the twelve task types and describes Reading and Listening as two-stage adaptive; the content page states that “As the test adapts, test time and items may vary.” The listed 120 is the blueprint total, not a promise about one sitting.S1, S2
A wider comparison of circulating claims against what ETS publishes is kept on the myth-check page.
9.Sources
- S1
ETS. TOEFL iBT Test: 2026 Update — Test Blueprint and Specifications Document. Read .
Downloaded and read through the PDF text layer (pdftotext, default column order). ETS serves the same document from a second path, linked as “View the TOEFL iBT Specification Document” on the test-content page: /content/dam/ets-org/pdfs/toefl/toefl-ibt-test-specifications-2026.pdf.
The document carries no revision date. It carries instead a Note on Specification Finalization, quoted in full below, which is the only statement it makes about how long its values hold.
- S2
ETS. TOEFL iBT test content and structure. Read .
Read from the raw HTML served to a client that runs no JavaScript — the same bytes an answer engine receives. The section table, the scoring notes and the FAQ answers quoted here are all present in that response.
- S3
ETS. Understanding your TOEFL iBT scores. Read .
Read from the raw HTML served to a client that runs no JavaScript.
10.Citing this page
This page is free to quote and link. If it is useful, cite it as:
Reach120 research team. “The TOEFL iBT format from January 2026: a sourced reference.” Reach120, . https://www.reach120.com/toefl-format-reference
The same rows, machine-readable
Every fact on this page is also published as a document with one entry per fact, each carrying its source URL, the date that source was read, and a permanent link to the row above: /citable-facts.json (37 entries). The anchors on the tables above are permanent, so a link to one row keeps resolving to that row.
Licence reach120-citation-1.0: cite, quote, link, print. Required: attribution, link-to-source-page. Attribute as {title} — Reach120. {url}. Full terms at /cite.
Reviewed . Next review due , or immediately on any change to the sources above. If a statement here no longer matches its source, the source is right and this page is wrong — tell us at /contact.