Automatically marked CEFR Placement Tests: Designed for English language schools

We rebuilt our placement test from the ground up. Here's everything that's new.

I've been teaching English for fifteen years. I ran into the same problem every language school runs into every single week: a new group of students arrives, and there's a mad rush to place them.

That's why we built a placement test that automatically marks all their work. And today, we're releasing the most significant update to it since launch.

Here's what's new.

A placement test that fits how your school actually works

Most placement tests are built for universities: one intake, one deadline, one big cohort sitting in a hall. Language schools don't work like that. Students arrive every week, sometimes every day, and you need to know their level before they arrive, not after.

So the whole test is designed around that reality. You send a student a single link when they book. They take the test in their own time, no headache of scheduling Zoom calls for you. It takes them about 20–25 minutes and covers all four skills: speaking, reading, writing, and listening.

You’ll get a CEFR level for each skill and their overall level, and we tell you whether they're sitting at the top or bottom of that band. So if you’ve got a full B1 class and you're deciding who to bump up to B2 or down to A2, you're not guessing. You can see exactly how close to the line they are.

Built to stop cheating, not just detect it

An AI-marked test is only useful if you trust the result. So we went hard on this.

The test is timed, question by question, and there's no pausing to go and look something up. We've disabled copy-paste on the writing section specifically so students can't paste in an AI-generated answer, so what you get back is actually their writing. 

Question packs are varied so students sitting next to each other aren't looking at the same paper. Questions are adaptive too, so an A2-level student doesn’t have to trudge through dozens of C1 questions feeling horrible about themselves. 

And if something looks off, you'll know. We flag it if a test has been restarted multiple times, and we flag it if more than one voice shows up during the speaking section (for example, if their teacher is translating the questions for them). 

Higher completion rates, without the chasing

This one came directly from schools using the test: students starting a test, getting interrupted - a bad connection, a knock at the door, whatever - and never finishing. Previously, that student just sat there, blocked. 

Now, if a test isn’t completed, we email them automatically to remind them. And if they still haven't finished, you can can send a one-click reminder. There’s no more chasing up individual students on WhatsApp. 

See test results only when you want them 

This is probably the biggest change, and it came from watching how schools actually use the results. Almost every school we spoke to only really needs a student's placement result in the week they arrive, or the week before, when timetables get built.

So our new dashboard organises everything by start date. Students are grouped by the week they're due in, which means when you're building next week's timetable, everyone you need is already sitting together - not scattered across a list sorted by submission date or surname.

From there you can sort by name, group, or level, and adjust a student's start date if their plans change. And because we know student names don't always transfer cleanly between booking systems - especially names that don't use the Latin alphabet - we've added fuzzy search. You don't need the exact spelling to find who you're looking for.

At a glance, you can see each student's overall level alongside their individual skill levels, so you can immediately spot the ones with a spiky profile - strong speaker, weaker writer, or vice versa.

Export exactly what you want

We’ve improved the export to let you select whichever students you're looking at - say, everyone arriving next week - and choose exactly what to pull: scores, individual skill breakdowns, names and emails, strengths, areas for improvement, or full summaries.

Use it to build your timetable, or hand it straight to teachers so they know their students before day one - what they're good at, what they need support with, and a little about who they are as people. Teachers love this head start on planning their lessons.

Two specialist AI engines, both trained specifically for language learners

This is the part I'm proudest of, because it's the part that's hardest to get right.

There are two separate AI systems working behind the scenes. One scores the CEFR level itself. We didn't just point a general-purpose model at a test and hope - we had experienced English teachers listen to the same recordings and score them, and used that to train the engine on what a genuine B2, or A1, actually sounds like. It's been refined this way over several months, and it keeps getting more accurate as it's checked against real teacher judgement.

The second engine analyses the content of what a student actually says or writes — their strengths, the language they're using well, and where they need to improve. It's built specifically to give feedback that's useful to a language teacher, not generic commentary.

You're not just trusting a black box, either. Teachers can listen to the original speaking recordings themselves, and writing samples can be downloaded as a CSV - so you can drop them straight into whatever LMS or system your school already runs, however it's set up.

Try it for free

We're offering 50 free tests to try the new version. No commitment, no card details — just a chance to see how much time it saves your team before your next intake arrives.

— Francis