OET Speaking Criteria 2026: How the Bands Work

🦈 The OET speaking criteria reward one thing above all: being easy to listen to while you run a real patient conversation. SpeakShark gives you daily spoken-English practice with phoneme-level pronunciation feedback while you speak, plus an AI teacher that remembers the mistakes you keep repeating between sessions. Free tier, no card. Start a free speaking session →

The OET Speaking sub-test is marked on nine criteria in two groups: four linguistic criteria scored 0–6 each, and five clinical communication criteria scored 0–3 each. To reach grade B — the 350 mark almost every regulator asks for — OET states that test-takers "will have achieved predominantly scores of 5 out of 6 on each linguistic criteria and 2 out of 3 for the clinical communication criteria." That single sentence, published in OET's own Speaking sub-test FAQ, is the most useful number in the whole exam and almost nobody quotes it. This guide takes it apart: what a 5 looks like versus a 6, what "competent use" means on a clinical criterion, how the raw marks turn into a 0–500 score, and — honestly — which parts of the conversion OET does not publish at all.

In this guide: criteria at a glance · the format the criteria are applied to · the four linguistic criteria · the clinical communication criteria · what a B actually requires · how the bands map · training each criterion out loud · how we checked this · when criteria study is the wrong move · FAQ

OET speaking criteria at a glance

Two things in that table surprise most candidates. First, more than half of the named criteria are communication criteria, not language criteria — OET is grading how you run a consultation, not just how well you speak. Second, the actual arithmetic that turns nine criterion scores into a number like 370 is not public. We come back to that in the B section, because it changes how you should read every OET score calculator online.

The speaking sub-test format the criteria are applied to

Criteria only make sense against the thing being marked. Per OET's own Speaking preparation page, the sub-test runs approximately 20 minutes and contains two profession-specific role plays of about 5 minutes each. OET's test-day guide adds the rest of the choreography:

• The session starts with a warm-up to help you feel comfortable — and OET says plainly, "This part is not assessed."
• Before each role play you get your role card on screen and three minutes to read the task, take notes and ask the interlocutor about unfamiliar words.
• You play your own profession; the interlocutor plays the patient or their carer.
• The whole session is recorded and marked afterwards by trained assessors — the person you are talking to is not scoring you.
• Delivered at home, OET says the sub-test "takes approximately 20–30 minutes, including all security checks."

Two consequences follow directly. Because the warm-up is unassessed, spending it settling your voice and breathing is free — our sibling guide to OET speaking warm-up questions treats it as a rehearsal room rather than a test. And because there are only about ten assessed minutes in total, every criterion has to be visible fast: an assessor cannot award relationship-building marks for empathy you never voiced.

There is also a rule that quietly protects you. OET's FAQ says there is no penalty for not completing all the elements on the role card — but "the more elements of the role-play you cover, the more evidence you are likely to give of your ability to communicate in spoken English." The criteria are evidence-based. Silence scores nothing; attempts score something.

The four linguistic criteria and what each band means

OET publishes full band descriptors in its Speaking assessment criteria and level descriptors sheet. Here are bands 6, 5 and 4 condensed — 5 is your B target, 4 is where B slips away.

Read down the band 5 column and one word recurs: occasional. That is the whole difference between a 5 and a 4. A band 5 speaker still makes errors — OET says so in writing — but those errors do not make the listener work. A band 4 speaker makes the listener work "at times." Nobody is asking you to be flawless.

The intelligibility row deserves its own paragraph, because it is the criterion candidates most often misread. OET's assessment criteria glossary states: "While L1 accent is to be expected in even the most able candidate, the main point to consider is the extent to which this causes strain for the listener. In many cases, accent poses no impediment to communication." The glossary then lists what assessors actually check — clear final consonants, recognisable vowels, correct word stress, voice projection without mumbling, selective stress to enhance meaning, and natural English sentence rhythm.

That list is a training plan, not a personality test. It is also why OET speaking for nurses and OET speaking for doctors both push the same drill: read a clinical explanation aloud and listen back only for final consonants and word stress.

Fluency is scored on rate and flow. The glossary asks whether you speak "at a normal rate (not too fast or too slow)" and "continuously and smoothly, with pauses or hesitations that are situationally appropriate, rather than a sign of searching for words." Note the distinction: a deliberate pause after bad news is appropriate hesitation. A pause while you hunt for the word "swelling" is not. Same silence, opposite score.

Appropriateness of language is where medical vocabulary can actively hurt you. The glossary asks whether you use "expressions comprehensible to a lay person in explaining technical procedures or medical conditions." Saying oedema to a frightened relative is not an advanced answer — it is a lapse.

Resources of grammar and expression looks for range plus the ability to paraphrase, "maintain longer utterances rather than single sentences, with appropriate use of cohesive devices," and use idiom accurately. Short, safe sentences protect accuracy but cap this criterion.

The OET speaking clinical communication criteria

This is the half of the mark scheme that has nothing to do with your English level and everything to do with how you run the conversation. Five criteria, each scored 0–3, each with named sub-indicators listed in OET's official descriptors sheet.

The scoring labels are blunt and worth memorising: 3 = adept use, 2 = competent use, 1 = partially effective use, 0 = ineffective use.

Three details in that grid change scores more than any vocabulary list.

D3 is a negative indicator. It is the only criterion phrased as something not to do: no compound questions, no leading questions. OET's glossary gives the shape — a compound question is "have you ever had chest pain or felt short of breath?", and a leading question is "You haven't had any ankle swelling?" Under pressure, second-language speakers stack questions because a single question feels too short. It is not too short. It is the scored behaviour.

E4 has a named better answer. The glossary says checking understanding with "does that make sense?" is weak, because "many patients will say 'yes' even though they are still unsure." The stronger move it names is patient restatement — asking the patient to say the plan back to you. That is the teach-back habit we drill in English speaking practice for nurses, and it is worth building until it is automatic.

Empathy is a scored behaviour, not a feeling. A4 asks for empathy "communicated back to the patient in a supportive way." The glossary notes that even on audio alone, silence and voice tone count — but "verbal empathy makes this more explicit by specifically naming and appreciating the patient's emotions." Naming the emotion out loud is the difference between a 1 and a 2. If your role play involves difficult news, OET speaking and breaking bad news works through the pacing.

And one reassurance, straight from OET's clinical communication guide: "The clinical communication criteria are language-based. You will not be evaluated on your medical knowledge." You are not being tested on whether your treatment plan is correct.

Reading these criteria takes ten minutes. Making them automatic while a simulated patient interrupts you takes weeks of talking. That gap is exactly what a conversational AI partner is for — and SpeakShark's Role Play mode will improvise back at you instead of reading a script. Run your first role play free →

What a B actually requires

Here is the honest, fully-sourced answer to the question that brought most people to this page.

1. The number. Grade B is a scale score of 350 or above, out of 500. That is the level the UK's Nursing and Midwifery Council requires in listening, reading and speaking, with C+ (300 or above) accepted in writing.

2. The criterion profile behind it. OET's Speaking FAQ and its linguistic criteria guide say the same thing in the same words: "Test-takers securing grade B (350) will have achieved predominantly scores of 5 out of 6 on each linguistic criteria and 2 out of 3 for the clinical communication criteria." So the target profile is roughly 5, 5, 5, 5 on the linguistic side and 2, 2, 2, 2, 2 on the clinical side. Not perfection — competence, repeated across nine boxes.

3. What OET does not publish. The conversion is the gap. The scoring page says only that "scores for each criterion are combined and converted to the 500 scale." OET does not publish the weighting between the two criteria groups, nor a lookup table from raw marks to scale score. Contrast that with Reading and Listening, where OET does give a raw benchmark: those papers have 42 questions, and "test-takers awarded Grade B (scale score of 350) typically achieve at least 30 marks." No equivalent raw threshold exists for Speaking.

The practical consequence: every OET speaking score calculator you find online is reverse-engineered from candidate reports, not from a published formula. Use them for direction, never for a decision. And note the word OET chose — predominantly. It implies a B is possible with one criterion below 5, and impossible to guarantee with several. Aim for the profile, not for a single heroic criterion.

How OET speaking bands map to grades, CEFR and IELTS

OET publishes both benchmarking tables itself. These are OET's own equivalences, taken from its Understanding your score sheet and its results-benchmarking tables:

Two readings of that table are worth taking away.

• The B band is wide. 350 and 440 are both a B, but they are a full IELTS half-band apart. If your regulator asks for "B", 350 is enough. If you are aiming to be comfortable rather than borderline, aim at 400.
• C+ is not a near-miss of B on the criteria. C+ sits at CEFR B2 while B sits at C1 — a whole CEFR level. Candidates stuck at 300–340 are usually not one lucky day away from 350; they are one criterion band away, most often on intelligibility or fluency.

If you have never had your speaking level measured at all, get a baseline before you book anything — our free English speaking level test takes minutes, and it is a better starting point than a practice exam that assumes you are already close.

How to train each criterion out loud

The criteria are a checklist of behaviours, which means they are trainable by rehearsal rather than revision. A four-week structure that maps one-to-one onto the mark scheme:

• Week 1 — intelligibility. Record yourself explaining one treatment or discharge plan aloud, daily. Listen back for exactly three things the glossary names: final consonants, word stress, and whether your voice projects or trails off. Fix the two or three words that repeatedly blur; do not attempt an accent change.
• Week 2 — fluency. Speak for five unbroken minutes a day on a clinical topic. The target is not speed, it is the absence of searching pauses. Count your fillers on playback; halve them the next day.
• Week 3 — structure and information gathering. Run role plays where you consciously signpost ("First I'd like to hear how you've been, then we'll talk about the medication"), ask one question at a time, and summarise before each topic change. Ban compound questions for the entire week.
• Week 4 — information giving and empathy. Practise chunk-and-check: two sentences, pause, check. End every explanation with a restatement request. Name one emotion out loud in every rehearsal.

Three things make this work faster than reading model answers. You need a partner who responds unpredictably, so you practise recovering rather than reciting — our guide to English role-play scenarios explains why unscripted beats scripted. You need pronunciation feedback at the sound level, because "work on intelligibility" is not actionable but "your final /d/ disappears on prescribed" is. And you need continuity, so week 3 does not restart from zero.

That last point is where SpeakShark is genuinely different: its AI teacher carries your recurring mistakes, your level and your topics between sessions, so the app already knows what week 1 exposed when week 4 begins. No other conversational practice app in this niche does that — most greet you as a stranger every morning. The free tier gives three 5-minute sessions a day with teacher Sarah, 64 topics and no card; Premium is $10/month or $69/year (about $5.75/month) for 10-minute sessions with unlimited turns, four teachers across American, British, Australian and Canadian English, 320 topics and advanced feedback — see pricing. If a British-English regulator is your destination, rehearsing with the British teacher is the obvious move. For the role-play format specifically, our companion piece on OET speaking role-play cards walks through original practice scenarios.

How we checked these criteria

Every criterion name, band descriptor and score threshold above comes from a page or PDF we opened in July 2026, and each is linked so you can verify it.

Primary sources, all published by OET: the Results and scoring page for the 0–500 scale in ten-point increments, the statement that linguistic criteria are "marked out of 6" and clinical communication criteria "marked out of 3", the combination statement, and the Reading/Listening raw benchmark of 30 marks out of 42; the Speaking preparation page for the ~20-minute format, two 5-minute role plays, two assessors, the no-penalty rule and the grade B criterion profile; the Speaking assessment criteria and level descriptors PDF for every band 0–6 descriptor and the full A1–E5 indicator list with the 3/2/1/0 scoring labels; the assessment criteria glossary PDF for the definitions, the accent statement and the compound/leading question examples; the Ultimate Guide to OET Speaking Part 1: Linguistic for "0–6", "0–3" and "weighted equally"; the Part 2: Clinical Communication guide for the medical-knowledge disclaimer; the Understanding your score sheet plus OET's published CEFR and IELTS benchmarking tables for the grade bands; and OET's test-day guide for the unassessed warm-up, three-minute preparation and the fact that the interlocutor does not assess you.

Regulator source: the NMC's accepted English language tests page for the B (350 or above) requirement in speaking and the two-year validity window.

What we deliberately did not do. We did not reproduce any real OET role-play card, warm-up question or assessor transcript — the example phrasings on this page are original, written for rehearsal. We did not publish a raw-mark-to-scale conversion table for Speaking, because OET does not publish one and inventing one would turn guesswork into apparent policy. And we did not quote a "typical" criterion score for any grade other than B, because B is the only profile OET states publicly.

SpeakShark is a speaking-improvement tool, not an exam-prep provider, and is not affiliated with or endorsed by OET, Cambridge Boxhill Language Assessment, the NMC, or any other exam board, regulator or employer named on this page.

When studying the criteria is the wrong move

An honest guide has to say where this stops helping.

If you are below CEFR B2, criteria study will not move your score. The mark scheme describes behaviours you can only perform once the language is automatic. A candidate who is still assembling sentences cannot simultaneously signpost, chunk and check. Build raw speaking hours first — daily unscripted conversation for a few months — then come back to the criteria. If nerves are the blocker rather than vocabulary, start with overcoming the fear of speaking English.

If one specific sound is your problem, use a drill tool. Isolated-phoneme apps with mouth-position graphics do a narrow job well. Conversational practice — ours included — is the wrong shape for retraining a single stubborn consonant from scratch.

If you need the exam format itself, go to OET. Sample tests, real role-play cards and assessor commentary come from the exam body. We will not reproduce them, and any site handing you "real OET cards" is a reason to be suspicious, not relieved.

If your regulator accepts another test, check whether it suits you better. OET's clinical communication criteria reward warm, patient-centred conversation. If that is your strength, OET is the friendlier route. If you prefer a machine-scored, structured format with no interlocutor, our PTE Academic speaking guide covers the alternative — though not every regulator accepts it, and the NMC currently accepts only IELTS Academic and OET.

And if you have already passed, keep going. A B gets you registered. The first night shift asks for the same nine behaviours with real consequences attached — which is the argument for treating this criteria list as a career checklist rather than an exam one.

FAQ

Nine criteria, ten assessed minutes, and one honest summary: OET is not asking you to sound native. It is asking you to be easy to listen to while you run a patient conversation properly. Both halves of that are rehearsable — and the only way to rehearse them is out loud, most days, with something that answers back.

Start your free daily speaking session on SpeakShark → — three sessions a day, no card, pronunciation feedback while you speak, and a teacher that remembers what you got wrong yesterday.

All OET criteria, band descriptors, score bands and format details verified in July 2026 from the official OET pages and PDFs linked above. NMC requirements verified on the official NMC page. Requirements change — always confirm the current rules with OET and with your own regulator before booking.