Why we built OET Live

OET Live started, like most useful things, with a friend in a tight spot.

She was a senior ICU nurse from Manila, six years in the unit, fluent in English in every sense that mattered for her job. She'd been offered a role in a London teaching hospital. She had three months. She needed Band B in every OET sub-test.

She passed Listening on the first try. She passed Reading. She passed Writing on the second sitting. Speaking is where she got stuck. She scored C+ three times in a row.

We sat down to figure out what was happening, and the story was unsurprisingly nuanced. Her English at the bedside was fluent. She could de-escalate a frightened family member, hand over to a registrar, and write a discharge summary. But the OET Speaking sub-test was asking her to do something her job rarely required: produce a structured, five-minute, examiner-aware role-play to a script she had read three minutes earlier, while being graded on nine separate criteria at once.

Her score reports said C+. They didn't say where the marks went. The examiner had decided, opaquely, that her understanding the patient perspective was 3/5. Or her providing structure was 3/5. Or both. There was no way to tell from the report.

What she actually needed was straightforward and not available:

1. A patient who would talk back to her
2. A way to know, immediately, which criterion she was losing
3. Practice loops that ran on her schedule, not a tutor's

That's not really a tutor's job. A good OET tutor is expensive and rare. Their hour is too valuable to spend playing patient. Their feedback is usually qualitative and post-hoc — "you could have shown more empathy" — without specifying where in the transcript the empathy was missing.

This is the gap we built OET Live to fill.

What's actually different

There are plenty of OET prep tools. Most of them are:

• Static practice cards (no live partner)
• Generic IELTS apps with a clinical wrapper (no real OET rubric)
• Marketplaces matching you to human tutors (good, but expensive and scheduling-bound)

What we wanted was an AI patient that stays in character, listens to what you actually said, and scores you on the official OET rubric at the end. That meant building three things:

1. An AI patient on live, low-latency audio. This wasn't possible a year ago. The latest generation of real-time audio AI is the breakthrough — it pauses naturally, can interject, and responds to the actual content of your speech rather than a transcription delayed by a second. The result is conversational tempo that doesn't break the practice illusion.
2. A scoring engine calibrated to the 9 OET criteria. Not the 4 IELTS Speaking criteria. Not "general English". The exact 9 the OET rubric uses, scored independently, with reasoning. We started by writing rubric prompts, then graded our own examples, then compared model scores against examiner scores on calibration sets, then tuned. The same way you'd train a junior examiner.
3. A case bank that doesn't run out. We generate cases against profession-specific topic schemas — what conditions does this profession see, what tasks does this scenario type require — then audit every case against the same rubric used to score performance. The bank currently has 11,000+ cases across 12 professions. That's enough for a candidate to never see the same case twice during their study window.

Who it's for

OET Live is built for one specific person: a healthcare professional who works in their second language, has three to six months to clear OET Speaking, and is sick of not knowing which criterion is dragging their band down.

If that's you, join the waitlist.

What we're trying not to do

We're not trying to replace a human OET coach. The best human coaches are great. What we want is to be the practice ground between coach sessions — the thing you do at 11pm on a Tuesday after your shift, when no one is available to do a role-play with you.

We're also not trying to make OET easier. The test is what it is, and the standard exists for a reason — clinical safety is downstream of clear communication. What we want is to make the path to OET Speaking proficiency less opaque, less expensive, and less dependent on whether you have a friend who happens to be a native English speaker.

What's next

Phase 1 is the iPhone app + the website. Phase 2 is per-profession landing pages, more languages, and Apple Sign in with Apple. Phase 3 — if we make it that far — is the Writing sub-test, where similar gaps exist.

If you want to follow the work as we ship, the blog is the operating record. We will write about methodology, about pronunciation engines, about how the case bank evolves, and about specific lessons from candidates.

Join the waitlist for TestFlight access. We'll see you in the next post.