Felix, our AI marker, as your patient
You prepare with the candidate card for three minutes, then talk to Felix for five. He plays the patient, relative or carer on the card, and afterwards marks what you said against the criteria that can be judged from a transcript.
Felix does not assess Intelligibility, your pronunciation and intonation, because he works from a transcript and cannot hear you. His report is an estimate, not an official result, and never a score out of 500. The estimate assumes your Intelligibility is at B level: for a B on test day you would also need at least 5 of 6 for it.
The card as you first see it, and the role-play under way with your turns typed. Tap one to read it.
Speaking — Role-play card · Three minutes to prepareNursing
Speaking — With the patient · Five minutesNursing
The same shape as the test: a card, three minutes, then the conversation.
The candidate card says who you are, who you are talking to and what to do: find out, explain, reassure, resist a request. Make notes; you may use them while you talk.
Speak or type. Felix plays the other side from his own card: what he knows, how he feels, what he wants. He ends the conversation at five minutes.
You get the criteria Felix could judge, with the words he found them in, the tasks on your card that you did and did not do, and a model dialogue to compare with.
He follows the role-player’s card.
Information comes when you ask for it.
Ask two questions in one breath and he answers one of them.
Somewhere in the conversation he says something that is really a concern. Catching it, and naming the feeling, is what the Understanding the patient’s perspective criterion is about.
He accepts a plan after a clear explanation, not before, and he does not pretend to understand when you have used a word he would not know.
The OET marks nine criteria in a role-play. Felix can judge eight of them from a transcript.
Felix reports a profile of these and an estimated grade. The OET’s own conversion to a score is not public.
These are things you can do on purpose, and Felix looks for each of them in your words.
“Can you tell me what happened?” before “Did it hurt?”. Closed questions are for confirming, once you have heard the story.
“Have you had chest pain or shortness of breath?” asks two things and suggests the answer. Split it.
Give a little information, stop, and ask the patient to say it back in their own words. “Does that make sense?” checks little.
“Now I would like to talk about what happens next.” A signpost costs five words and tells the patient where you are.
When a patient says “I’m just so worried”, name it and respond before you carry on with the plan.
If the card asks you to resist a request, for example for a diagnosis that is not your role, giving in is unsafe practice, and Felix reports it separately.
You can speak or type. If you speak, your browser’s own speech recognition turns your words into text: depending on the browser, that runs on your device or on its maker’s servers, and it is not our service. We never receive or store audio, only the text.
Speech recognition can mishear an accent. If the transcript is wrong, type instead; Felix marks the transcript, not your voice.
Pick a role-play for your profession. What is free, and what needs a pass, is on one page.
MedicalCertPrep is an independent preparation site. We are not affiliated with or endorsed by Cambridge Boxhill Language Assessment or the OET. ‘OET’ is a trademark of its owner and is used here only to describe the test our practice material prepares you for. Marking by Felix, our AI marker, is an estimate, not an official result.