All 12 TOEFL 2026 task types explained, with how each one is scored

The 2026 TOEFL® iBT has four sections, Reading, Listening, Writing and Speaking, in that order, and twelve task types across them. Here they are in test order, with what is on screen, how each one is scored, and one tip that follows from the format.

Reading

Up to 50 questions in about 30 minutes; 35 count, and the rest are unscored tryout items you cannot tell apart. Your first module decides whether the second is easier or harder. The text stays on screen while you answer. Each question is 1 point, marked automatically, and scores are equated across modules.

Complete the Words

One paragraph of about 60 to 90 words. The first sentence is whole; after it, the second half of every second word is missing, one slot per letter, until there are 10 gaps. 30 items.

Marked automatically; only the exact letters score.

Tip: the slots give the word's length, so a word that fits the meaning but not the count is wrong. Check the ending (-s, -ed) too.

Read in Daily Life

A text of 15 to 150 words: an email, a text-message chain, a notice, a menu. It stays beside the questions, 2 for a short text and 3 for a longer one. From 5 to 15 questions, varying with your path.

Marked automatically.

Tip: wrong options for a detail question are other times, dates and names from the same text, so check that your choice answers the question asked.

Read an Academic Passage

About 200 words in three paragraphs, with 5 questions: a word's meaning in context, a detail, an inference, why the author mentions something, how two paragraphs relate. Some use EXCEPT. From 5 to 15 questions, more in the harder second module.

Marked automatically.

Tip: in an EXCEPT question, three options are in the passage and the odd one is a small twist on what it says. The passage stays on screen, so check each option against it.

Listening

Up to 47 questions in about 29 minutes, 35 scored, adaptive like Reading. Every recording plays once, with a picture of the speakers; questions follow one at a time, each with its own clock and no going back. You can take notes. Each question is 1 point, marked automatically and equated.

Listen and Choose a Response

One person says one line of 4 to 12 words, which you hear and never see; four written replies are on screen from the start. From 15 to 19 items, depending on your path.

Marked automatically.

Tip: read the replies before the line plays, then distrust a reply that repeats a word from it: wrong options often reuse a word with another meaning, and the right one is often indirect ("Actually, …").

Listen to a Conversation

Two speakers in 4 to 6 turns of about 50 to 90 words. 2 questions a conversation, 10 in the section.

Marked automatically.

Tip: almost every conversation asks what a speaker will do next, and the last turn sets it up. Note the final line.

Listen to an Announcement

One speaker, about 40 to 70 words, in a classroom or on the campus radio: schedules, room changes, rules, events. 2 questions each; 6 to 10 in the section, none in the harder second module.

Marked automatically.

Tip: one question usually asks what the speaker wants listeners to do, answered by a short action ("Register online"). Note the instruction near the end.

Listen to an Academic Talk

A professor or a podcast host speaks for 175 to 250 words; no background knowledge needed. 4 questions a talk: the main topic, a point made, why something is mentioned, what is implied, what comes next. From 8 to 16 questions; the harder second module has two talks.

Marked automatically.

Tip: the answer to "What is the talk mainly about?" is a plain paraphrase of the topic; the wrong options are narrower points mentioned in passing.

Writing

Not adaptive: everyone gets the same 12 items in about 23 minutes. Build a Sentence counts for half the raw score; each written response, scored 0 to 5 on ETS's scoring guide by its AI scoring engine with human raters in the process, counts for a quarter.

Build a Sentence

Speaker A's line, then 4 to 7 blanks for B's reply with any fixed words and the end punctuation in place, and the word chunks you move into them. Every chunk is used once. 10 items, about 6 minutes.

Marked automatically, 1 or 0 each.

Tip: read the punctuation first. A question mark means B's reply is a question, as most are, so the auxiliary or question word goes first.

Write an Email

A short scenario, then "Write an email to [person]" with three bullets that each start with a verb. To and Subject are already filled. 1 item, 7 minutes.

Scored 0 to 5. A 5 has elaboration that supports the purpose, consistent social conventions and almost no errors.

Tip: the three bullets ask for different things, usually a description, a problem and a request. Give each its own short paragraph, in the register the scenario sets.

Write for an Academic Discussion

A professor's post ending in a direct question, two students' posts taking different positions, and "An effective response will contain at least 100 words". 1 item, 10 minutes.

Scored 0 to 5: a 5 is a clear, relevant contribution with well-elaborated explanations and examples and almost no errors.

Tip: the students are there to be built on. Agreeing with one and repeating the reason is a weak contribution; add a reason neither gave. This task is unchanged from the old test.

Speaking

Linear too: 11 items in about 8 minutes, no preparation time, each prompt played once before a clock counts down. Every item is scored 0 to 5 by ETS's AI scoring engine, with human ratings in the process.

Listen and Repeat

An illustration of a place such as a library; you hear each sentence once and repeat it, and the text is never shown. 7 sentences, growing from about 5 words to about 14.

Scored 0 to 5 by an engine that compares your words with the sentence and measures fluency and clarity: 5 is exact, 4 keeps the meaning with a small word wrong, 2 has a large part missing.

Tip: a dropped function word costs one point and stopping halfway costs three, so finish the sentence with the meaning you kept.

Take an Interview

An interviewer on video asks 4 questions in a set situation, such as a research study. They climb from your own experience to a choice, an opinion, then a broad social question.

Scored 0 to 5 on ETS's scoring guide for topic development, delivery (pace, pauses, intonation) and language use. An answer made mostly of the question's own words sits at 2.

Tip: each turn opens with an acknowledgement ("I see") and the question comes at the end. Listen to the whole turn, answer in your first sentence, then give one reason and one example.

Which task types to practise first

Take the free mock first: the report names your weakest task types. Then practise by section, Reading, Listening, Writing and Speaking; the first set of each task type is free. The one-minute Speaking check plays a Listen and Repeat sentence once and shows which words came through and your pace.

Are all 12 task types on every test?

Yes, but Reading and Listening adapt, so counts from some types change with your second module. The easier Reading module has Complete the Words and Daily Life texts; the harder has academic passages instead.

Which task types are scored by a person?

Reading, Listening and Build a Sentence are marked automatically. Written responses and Speaking are scored by ETS's AI scoring engine; ETS says the engines are trained on human ratings and human raters provide oversight.

Do the old Speaking tasks still exist?

No. The old independent and integrated tasks are gone, and nothing in Listen and Repeat or Take an Interview asks you to summarise a reading or a lecture.

Skorway's mock reports band estimates on the same 1–6 scale. They are practice estimates, not official scores. TOEFL® is a registered trademark of ETS. Skorway is not endorsed or approved by ETS.