Free to start — no card
TOEFL Speaking practice
Eleven items on the published windows — seven sentences repeated back and measured word by word, then four interview questions with ten seconds to think and forty-five to answer.
Format checked against ETS on .
The section we run
Eleven items, about eight minutes. This is the Speaking section ETS has set since 21 January 2026. The four integrated tasks the old paper was known for are not in it, and a tool still drilling them is drilling a section nobody sits.
Both spoken task types
- Listen and Repeat — a sentence is played and you say it back. No preparation; it begins, and you get 15 seconds. Seven items of the eleven.
- Take an Interview — four questions about ordinary life, work and opinions. 10 seconds to prepare, 45 to answer, each.
Ten seconds is not enough to plan an answer, and that is the design. The task rewards starting immediately with a position and one developed reason — a rehearsed introduction spends a quarter of your time saying nothing.
Speaking is linear, so a fixed form is the delivery
Reading and Listening adapt on the day and our fixed-form rehearsal of them is an approximation. Speaking is not adaptive: ETS states that every candidate on a form receives the same set of tasks. So eleven items in this order, on these windows, is not a stand-in for the real section — it is how the real section runs.
One item is measured, not judged
Listen and Repeat has a right answer, which almost nothing spoken does. Your words are lined up against the sentence you were played and what survived is counted — in sequence, so reading it backwards earns nothing and dropping a single early word costs one word rather than the whole item. The feedback names what you left out, which a band on its own cannot.
A digit never appears in a repeat sentence. “14” is spoken as “fourteen” and you would say the word, so the alignment would count a miss on a repeat that was perfect — marking you down for our script rather than for your speech. The generator refuses those sentences.
The rest is marked against the descriptors
Your interview answers are transcribed and marked criterion by criterion, with the reason given rather than a bare number.
Said plainly: a transcript is strong evidence for vocabulary and grammar, and weaker evidence for pronunciation than a human examiner listening to you. Transcribers also tend to correct — hearing an unclear word and writing the expected one — which flatters a candidate. That is the safe direction for practice and a reason to treat the delivery score as the least certain part.
Questions, answered
What are the TOEFL Speaking tasks in 2026?
Two, and eleven items between them. Listen and Repeat plays you a sentence and asks you to say it back — seven of those. Take an Interview asks you four questions about ordinary life and opinions, ten seconds to think and forty-five to answer each. The four integrated tasks that used to define the section were retired with the rest of the old format on 20 January 2026.
How long do I get for TOEFL Take an Interview?
Ten seconds to prepare and forty-five seconds to answer, four times. Ten seconds is not enough to plan an answer, and that is the design: the task rewards starting immediately with a position and one developed reason. A rehearsed introduction spends a quarter of your time saying nothing.
How is TOEFL Listen and Repeat scored?
By measurement rather than judgement. It is the one spoken item with a right answer, so your words are lined up against the sentence you were played and what survived is counted — in order, so reading it backwards earns nothing and dropping one early word costs one word. The feedback names what you left out. A digit is never put in the sentence, because “14” is spoken as “fourteen” and you would be marked down for our script.
How is TOEFL Speaking marked without a human examiner?
Your recording is transcribed and marked against the published descriptors, criterion by criterion. Listen and Repeat is measured against the sentence instead. Delivery and language use are judged from the transcript, which is strong evidence for vocabulary and grammar and weaker evidence for pronunciation than a person listening to you.
Is this the current TOEFL Speaking format?
Yes. ETS redesigned the TOEFL iBT on 21 January 2026 and this is the redesigned Speaking section: eleven items in about eight minutes across the two task types above, reported as a 1 to 6 band rather than a score out of 30. ETS also reports a comparable 0 to 120 total until January 2028.
Free to start — four speaking evaluations a month, which is one whole Speaking section, because each of its four interview answers is marked. No card.
The rest of the paper
TOEFL Reading, Listening and Writing have their own pages, and the full TOEFL practice test covers all four. Add them up on the TOEFL score calculator.