What a TOEFL mock test should include in 2026, and what the free practice tests leave out

The short answer

A TOEFL mock test, an exam simulation, a full-length practice test: whatever the name, it has one job, to tell you the band you would get if you sat the real test today, and which section is holding it down. To do that for the test as it has run since 21 January 2026, it needs eight things: the 2026 format with its 12 task types; the real section timers, running without pause; audio that plays once, with no transcript; a second Reading part and a second Listening part chosen by your first; the 1 to 6 band scale; Speaking and Writing that are actually scored; a report that names task types, not only bands; and a microphone check with recording timers for Speaking. Most free tests have two or three of the eight. ETS's free full-length tests are paper forms with answer keys for Reading, Listening and Build a Sentence only; its scored online test costs US$49 for one attempt.

A mock is not practice

Practice answers "how do I do this task". A mock answers "what would I score". The two need different things. A practice set can stop the clock, show the transcript and explain every answer as you go. A mock has to do what the real test does: run the sections in order, hold the timers, play each recording once and score the whole thing afterwards, so that the number at the end means something. The six-week plan uses both, and this guide is about the second kind, because it is the one that is hard to find.

The eight things to check

1. The 2026 format, with the 2026 task names

The test changed completely in January 2026: four sections in the order Reading, Listening, Writing, Speaking; 12 task types; up to 50 Reading and 47 Listening questions, 12 Writing tasks and 11 Speaking items; about 90 minutes in all. The What changed guide has the full list. The quick test of any mock is its task names. If it has Complete the Words, Build a Sentence, Write an Email, Listen and Repeat and Take an Interview, it is the 2026 test. If it has three or four reading passages with ten questions each, lectures of five minutes, an integrated Speaking task or an independent essay, it is the old test, and the score it gives you is on a scale the test now reports only as an estimate beside the band.

2. The real timers, running without pause

The sections are timed as a whole, and the clock does not stop for a drink or a phone call. Reading is about 30 minutes, Listening about 29, Writing about 23 and Speaking about 8; inside Writing, the email has 7 minutes and the academic discussion 10. A mock with the right questions but no clock, or a clock you can pause, measures your English and not your test. The easiest points to lose in Reading are lost to time, 50 questions in about 30 minutes being 36 seconds each, and that is exactly the loss an untimed test hides.

3. Audio that plays once, with no transcript

Everything you hear on the test, the Listening recordings and the Speaking prompts, plays once. There is no replay and no transcript, and in Listening the questions appear only after the recording ends. A mock that shows the transcript, lets you replay, or prints the Listening lines as text, cannot score your Listening, because reading a conversation and hearing it once are different skills. The same goes for Speaking: Listen and Repeat is a sentence you hear once and say back from memory, so a printed sentence is a different task.

4. A second part chosen by the first

Reading and Listening are two-stage adaptive. Your answers in the first part decide whether the second part is the easier set or the harder one, and the band is worked out from both. A fixed, one-path test cannot show you what the harder second part feels like, and it cannot convert your raw score the way the real test does, because the conversion depends on which path you took. The adaptive Reading guide explains the two paths.

5. The 1 to 6 band scale

Since January 2026 every section is reported as a band from 1 to 6 in half steps, and the overall band is the average of the four, rounded to the nearest half. The score report also gives a comparable 0 to 120 total for two years, for universities that have not updated their requirements. A mock that gives you a score out of 30 per section, or out of 120 only, is using the old scale, and converting it yourself through the concordance table is a rough estimate at best. The band score guide shows what each band means and who asks for it.

6. Speaking and Writing that are scored

This is where most free tests stop. Reading and Listening are easy to mark by key. Writing's Build a Sentence is too. The email, the academic discussion and all 11 Speaking items are each scored on a 0 to 5 scale, by an AI scoring engine and human raters together, and they make up half of your overall band. A mock that leaves them unscored, or gives you a sample answer to compare yourself against, gives you two bands out of four, and no overall. Ask how the scoring is done and how it was checked against real scores. If the answer is "a tutor will mark it within 48 hours", you have a marking service, not a mock.

7. A report that names task types

A band tells you where you are. It does not tell you what to do next week. Two candidates with a Listening 4 can have different problems: one loses the Choose a Response items, which are short and fast, and the other loses the academic talk, which is long and dense. A useful report breaks the section into its task types and names the ones that cost you the most, and for Speaking and Writing it shows you the answer with the mistakes marked, so that the next version is better and not merely longer.

8. A microphone check and recording timers

Speaking on the test is 11 recordings against a clock with no preparation time, into a microphone the test checks before you start. A mock that lets you type your Speaking answers, or record without a timer, or skips the check, is not testing the thing that goes wrong on the day: the recording that starts late, the sentence that runs out of time, the answer that trails off because the clock was invisible. Run the mock through the speakers and microphone you will use on test day, which for the Home Edition means no headset.

What the free tests leave out

The three ETS products on the preparation page, as they stand on 6 November 2026, against the eight points above.

	Free full-length practice tests 1 and 2	TOEFL Practice Online	Official Guide, 8th edition
Price	Free, after an email form	US$49 for one attempt	US$29, eBook
Format	2026, adapted for paper	2026	2 tests; the product page does not say which format
Timers	None; you time yourself	Timed	None
Audio	None; Listening lines and Speaking prompts are printed as text	Plays as on the test	Media for Listening and Speaking
Adaptive second part	No; both parts fixed	Yes	No
Band scale	No score; answer keys only	Section scores and an overall score	Answers and explanations
Speaking and Writing scored	No; keys for Build a Sentence only	Yes	No
Report by task type	No	No, and no answer review	Explanations per question

The free tests are good for one thing: they show you real test items at real length, and the task types guide uses them for exactly that. They are not a mock. There is no clock, no audio, no routing and no score, and the Speaking section is a list of printed questions. TOEFL Practice Online is a real mock, timed, adaptive and scored, and it is the one to take if you want an official number. It gives you that number and nothing else: no answer review, so you cannot see which questions you missed, and skipping a Speaking or Writing task can make the test non-scoreable.

Free tests from prep companies vary more than the official ones. Three checks take two minutes: look for Take an Interview and Build a Sentence in the task list, which tells you whether the test is the 2026 one; look for a visible timer on the first screen; and look for what happens to your Speaking recordings and your two Writing tasks after you submit them. A test that hides those three until you have given an email address usually has a reason.

How to take a mock so the band is real

• One sitting, in order. Reading, Listening, Writing, Speaking, with no pause between them. A mock split over two evenings scores higher than the test will, because fatigue is part of what the test measures.
• The room and the equipment you will test with. For the Home Edition, that is your own desk, speakers and an open microphone; for a centre, a quiet room and a headset.
• Listen, do not write. Notes are allowed, but the recordings are short, a 60-word conversation is over before a note is finished, and a mock is the place to find out whether you can do without them.
• Twice, at least. Once at the start, to get your band and your weakest section before you decide how many weeks you need. Once two or three weeks before the test date, to check that the work moved the band and that nothing new has gone wrong. A third, a few days before, is for timing and nerves, not for learning.

How to read the result

The band for each section is the first thing to look at, and the lowest one is where the next weeks go, because the overall band is an average and the weakest section moves it most. Read the gaps with care: the real test's measurement error is between 0.2 and 0.4 of a band per section, so a 4.5 and a 5 on two mocks a week apart may be the same ability on two days. A full band is real. Half a band in one section, twice in a row, is real. A half-band wobble between two sittings is noise, and chasing it is a waste of a week.

Take the free mock before you pay for one

Skorway's free mock is the full 2026 test built to the 2026 specification: four sections in order, 115 questions and tasks across the 12 task types, the real timers, Reading and Listening with a second part chosen by your first, and every recording played once with real voices. When you finish, Reading and Listening are marked by key and every Speaking and Writing answer is scored twice, independently, with a third pass when the two disagree. A few minutes later you have a band for every section, an overall band, a 0 to 120 estimate, and each Speaking and Writing answer marked up, explained and rewritten one band higher. The report names the three task types that cost you the most, and the rest of the site works on those. After every official test we ask our users for their real score and publish the comparison on the Accuracy page. One full mock is free, with the full report. No credit card.

Is there a free TOEFL mock test for the 2026 format?

ETS's two free full-length practice tests are in the 2026 format but are paper forms: no timers, no audio, no adaptive routing, and answer keys only for Reading, Listening and Build a Sentence. Skorway's free mock is the full 2026 test, timed and scored in all four sections.

How many mock tests should I take before the TOEFL?

Two at least: one before you plan, one two or three weeks before the test date. Between them, work on the weakest section with practice sets, not with more mocks. A third mock in the last week is for timing.

Is TOEFL Practice Online worth the US$49?

If you want an official number, yes. It is timed, adaptive and scored. You get section scores and an overall score and nothing else: no review of your answers, no feedback on Speaking or Writing. Take it after a free mock has told you what to fix, not before.

Does a mock test score match the real TOEFL score?

Close, if the mock is built to the real format and scored the same way, and no mock is exact. The real test itself has a measurement error of between 0.2 and 0.4 of a band per section. Expect a mock band to land within half a band of the official one, and treat the weakest section, not the decimal, as the information.

Correct for the TOEFL iBT as it runs on 6 November 2026. TOEFL® is a registered trademark of ETS. Skorway is not endorsed or approved by ETS.