These tests are not ranked by difficulty, whatever the forums say. They differ in who or what is doing the marking, and that single fact decides which of your weaknesses will show up in the score and which will pass unnoticed. A candidate who hesitates and thinks is expensive to a machine and cheap to a person. A candidate who is fluent and slightly off the question is the reverse.
So the choice is made twice. First against the list the institution published, which is not negotiable and takes ten minutes to check. Then, among whatever survives that list, against the kind of marking that is kindest to the way you actually perform.
The six that decide which one you sit
- choose by what the institution accepts, first and always. The best exam is the one your university or employer names.
- IELTS, TOEFL, Cambridge, PTE, the four that matter, and they differ more in format than in difficulty.
- band scores against a pass. IELTS and TOEFL report a score; Cambridge reports a pass at a level and the certificate does not expire.
- most scores expire after two years, which is a real constraint on when to sit it.
- the speaking test is with a person or a computer, and this is the single largest practical difference between them.
- the exam tests exam skills too. Familiarity with the format is worth a meaningful part of the score and is the cheapest preparation available.
A rater forms an impression, an engine takes a measurement
A human marker works from a rubric under time pressure, hears a couple of minutes of you, and arrives at a judgement that the rubric is then used to justify. That is not a criticism; it is what trained marking of that kind is, and it is consistent enough to have run for decades. It means the answer that addresses the question, sounds like a person, and takes a risk with a phrase will be rewarded, and a long silence can read as somebody thinking rather than somebody stuck.
An engine cannot form an impression. It extracts what it can measure reliably from the recording or the script, mostly delivery, range and accuracy, and compares those against a model built from very large numbers of previous candidates. An engine is strong on pace and pausing, it cannot be charmed, and it is much weaker on whether the answer was worth making, so a fluent, well-paced response that slides off the topic can score better than it deserves. The practical consequence is not that one is fairer. It is that computer scored, and what that changes about preparing and the human-marked papers reward different rehearsals, and preparing for one while sitting the other is a real and common way to lose half a band.
Two lists, kept by people who never speak to each other
The first decision is not which test suits you, it is which test is accepted for the thing you are doing, and there is a trap in that sentence. For anyone moving country, there are usually two acceptances, not one. The university or the employer keeps a list. The immigration authority keeps its own, often shorter, and the two are maintained by different bodies with no obligation to match. A test that satisfies the admissions office and not the visa route is a test you will sit twice, at your own expense, having prepared perfectly.
Check both lists, in writing, on the institution's own pages rather than on a preparation site, and check the date on the page. This is the entire content of which test your university or visa actually wants, and it is boring in a way that saves people several hundred pounds and three months a year.
The other structural difference worth knowing before you book is what expires. IELTS and TOEFL scores are normally treated as current for two years, and the Cambridge qualifications in B2 First, C1 Advanced, and the certificates that never expire do not expire at all. If the plan is a job application in four years rather than a visa next spring, that difference is worth more than half a band.
The score stops moving after the fourth practice test
This is where preparation stalls, and it is a specific and very common trap. Somebody buys a book of practice tests and works through them one after another, and the score climbs for the first two or three and then sits still for a month. Nothing is wrong with the person or the book. A practice test measures; it does not teach, and after the third one it has finished telling you what it knows about you.
What moves the number again is the ratio. For every hour of sitting a test, spend three on the specific thing it exposed, and do that work on material that is not a test: if the listening section keeps costing you the map or diagram questions, the answer is not another listening section but an hour with a recording, a pencil and no time limit. Test yourself monthly. Practise everything else in between, which is also the honest reading of the four papers, the band scores, and what examiners reward, since three of those four are ordinary skills and only the fourth is a format.
Who accepts what
The two acceptances, and the one conversation nobody has
A score is only worth the lists it appears on, and for anyone moving country there are two lists rather than one.
- You send a number. They rarely explain where it came from.
- No shared list, and neither one checks the other's.
- A test they will not take is a test you sit twice.
- You
- The university or employer
- The immigration authority
- YouPay once, sit once, and hope both lists agree
- The university or employerPublishes a minimum, and a floor per component
- The immigration authorityKeeps a shorter list, for its own reasons
- You — The university or employerYou send a number. They rarely explain where it came from.
- The university or employer — The immigration authorityNo shared list, and neither one checks the other's.
- The immigration authority — YouA test they will not take is a test you sit twice.
The marked side is the relationship you are not in, and it is the one that decides whether the money was well spent. Both lists are published, and reading them is the cheapest hour in this cluster.
What to prepare with, and what not to pay for
Every board publishes sample papers for its own test, free, in the current format. Start there and never anywhere else, because those are the only materials guaranteed to match what you will actually sit. They replaced the third-party preparation book, which reconstructs the format from memory and from the last edition, and which is always at least one revision behind. Free, in 2026, from all of them, and the free set is large enough to run the monthly test for a year.
For a machine-marked test the exception is worth the money. Pearson sells scored practice tests for PTE that run on the same engine as the real thing, and they are the only practice score that means anything, because no teacher and no third-party mock can reproduce what an engine counts. Not free, and deliberately so. If you are sitting the academic test most American universities ask for or the cheap fast one, and who accepts it, the same principle applies in a weaker form: the official practice is the one whose score you can trust.
The mistake: spreading the last month evenly
Most institutions publish an overall requirement and a minimum for each component, and most candidates read the first number and not the second. A candidate with an overall band above the requirement can still be refused for a single component that sits under a stated floor, and it happens constantly. In the final month the marginal hour does not belong to your weakest skill and it does not belong to your favourite. It belongs to whichever component is closest to falling below its own published minimum, which is usually writing, and which is a different question from what your English needs in general.
Which of the six is your question
If the lists still have to be checked, everything else waits. If they are checked and two tests survive, choose by marking rather than by reputation, and sit whichever one rewards how you actually perform under pressure. What none of this measures is the thing the number is standing in for, and the gap between them is worth keeping in view: where you are now, and what the next step actually contains is the same subject without a fee attached, and the descriptors there are free to read tonight.
The two papers that decide most results are the two that are hardest to practise alone. The written one is a format problem sitting on top of a real skill, and the real skill is emails, essays and reports that read like a native wrote them. The spoken one cannot be rehearsed by reading anything at all, which leaves the question of who you are going to practise with between now and the test date, and that is getting from correct sentences to an actual conversation.