Adaptive and short. Questions get harder as you get them right, so the test finds your level in about an hour instead of testing everyone on everything.
That adaptivity is the thing most guides skip, and it changes how you should prepare. You will not see every task the same number of times as the person next to you, and the test stops serving a type once it is confident of your level — so a short run of one task is information, not luck.
Short, adaptive, and served in bursts throughout the test rather than in one block. These are where the engine finds your level, which is why they get harder the moment you start getting them right — and why a run of easy-looking words is not the good news it feels like.
Decide whether each word is real English. Five seconds each.
Complete the missing letters of one word in a sentence.
Type the sentence you hear. Up to three plays.
Longer, single-clock tasks where the pressure is budgeting rather than knowledge. Interactive Reading gives you six questions on one shared timer, so time spent on the first is time taken from the last.
Restore the missing letters across a damaged passage.
One passage, six questions, one shared clock.
Follow a conversation, choose replies, then summarise what happened.
One image, one minute or ninety seconds, no prompt beyond the picture. These are the tasks candidates most often under-fill, and an unfinished response is scored as what it is.
Describe the picture in one minute.
Describe the picture aloud for up to ninety seconds.
Fixed-format rather than adaptive: there is one prompt, and its difficulty is not what is being manipulated. Interactive Speaking is the exception in this group — thirty-five seconds per answer, adaptive, and no second takes.
Read a prompt, then speak to it for up to ninety seconds.
A conversation. Thirty-five seconds per answer, no second takes.
Write for five minutes, then extend your answer for three more.
Last in the sitting, and different in kind from everything before them: these are scored AND sent on to the institutions you apply to. They are the only part of the test a human admissions officer will actually see.
Five minutes. Scored, and sent to the institutions you apply to.
Up to three minutes. Shared as video with your institutions.
There is no Reading section. Responses feed subscores directly, and several tasks feed more than one — Interactive Listening is marked for listening and, because the summary is typed, for writing as well. The overall score is the average of the four individual subscores, and the four integrated scores are averages of pairs.
The average of Reading and Writing.
The average of Reading and Listening.
The average of Listening and Speaking.
The average of Writing and Speaking.
A university that asks for 115–130 is asking about the overall figure, but many also set a floor on one subscore — which is why an average built on a weak speaking score can still fail a requirement it appears to clear.
What the scores mean — the 10–160 scale against CEFR, IELTS and TOEFL
Thirteen. They are interleaved through the sitting rather than grouped into sections, and the short adaptive ones arrive in bursts with other work between them.
The short items are — questions get harder as you answer correctly, which is how the test finds your level in about an hour. The extended responses and the two samples are fixed-format, because their difficulty is not what is being manipulated.
The test stops serving a type once it is confident of your level, so a short run of one task is information rather than luck. Two candidates will not see the same number of each item.
There are no sections. Responses feed the four subscores directly, and several tasks feed more than one — Interactive Listening is marked for listening and, because the summary is typed, for writing too.
They are scored like everything else, and they are also forwarded to the institutions you apply to. They are the only part of the test an admissions officer reads or watches directly.
Our Duolingo mock runs all 13 task types on the published timings, adapts the way the real test does, and returns the four subscores plus the integrated ones on the 10–160 scale. The first session is free.
This guide is written from the Duolingo English Test test format published by Duolingo, in our own words. For the official wording, see Duolingo's official test resources. MockLive AI is an independent practice tool and is not affiliated with Duolingo.