Why speaking is the hardest paper to prepare alone
Every other paper gives you a way to check yourself. Reading and Use of English have an answer key. Listening has a transcript. Writing at least leaves a text you can reread with fresh eyes tomorrow.
Speaking leaves nothing. You cannot hear your own hesitation while producing it, you cannot notice the structure you avoided, and you certainly cannot tell which of the five assessment scales is the one dragging your band down. Most candidates practise speaking by speaking more — which builds fluency and does very little for the specific thing that is wrong.
A teacher solves this. So does a study partner who knows the criteria. Both are also the two things most self-study candidates do not have on a Tuesday evening.
What an AI examiner can genuinely help with
| Works well | Why |
|---|---|
| The Part 2 long turn | A solo minute against a clock — no partner is involved in the real exam either, so nothing is being simulated away |
| Extended Part 4 answers | Position, reason, example. The shape is assessable from your answer alone |
| Speculation and hedging language | Whether you actually produced it, or fell back on describing, is visible in a transcript |
| Range and repetition | Machines are better than humans at noticing you used the same structure eleven times |
| Task achievement | Did you answer the printed question, or the one you wished had been asked |
The common thread: these are all things a transcript contains. Where the evidence is in what you said, feedback can be specific and immediate, and immediate is what makes it stick.
Try it nowRecord one long turn and read the feedbackThe first Part 2 drill is free, no account needed. Speak into the microphone or type.Try itWhat it cannot do
Worth being blunt about, because the limits decide how you should use it.
- It cannot rehearse Interactive Communication. That scale is about how you respond to another person — inviting a quiet partner in, disagreeing without ending the exchange, changing your mind because of what they said. There is no partner, so there is nothing to assess. For Parts 3 and 4, a human study partner remains the better practice.
- It cannot give you an official result. Any band you see is an indication against the public assessment scales, not a Cambridge grade, and no automated mark — ours included — replaces a trained examiner.
- It cannot reproduce exam-room pressure. Two strangers, one of them silently writing, is a specific experience. Nothing you do at home is that.
- It cannot fix what you never attempt. If you avoid complex structures, feedback will tell you the range was narrow — which is true and unhelpful until you push into them deliberately.
How to use it, part by part
- Part 12 minInterviewQuestions about you
- Part 24 minLong turnA minute each with two of three pictures
- Part 34 minCollaborative taskDecide something together
- Part 45 minDiscussionThe examiner opens the topic out
15 minutes · in pairs · two examiners
Part 2 — the highest-value target
The long turn is where solo practice maps most cleanly onto the real task. Record a full minute without stopping, then read the feedback for one thing only: did you speculate, or did you describe? That single distinction is the C1 delta on this part, and it is easy to check in a transcript.
Parts 3 and 4 — target the language, not the interaction
You cannot rehearse negotiating alone, but you can rehearse the phrases that make negotiation possible, and the extended-answer shape Part 4 rewards. Practise producing a position with a reason and a qualification on demand; the interactive half is what you save your study partner for.
Part 1 — least worth the time
Short personal answers are the part you can most reliably self-assess. Spend the effort elsewhere.
Fitting it into a week
Speaking rewards frequency over duration — fifteen minutes four times a week beats an hour on Sunday, because what you are training is automaticity. A workable shape:
- Twice a week: one Part 2 long turn, read the feedback, redo the same prompt immediately with the one fix applied. The redo is where the learning happens.
- Once a week: a Part 4 question, aiming for a three-sentence answer with a qualification in it.
- Once a fortnight, with a person if you have one: a full Part 3, because that is the gap AI leaves.
If you cannot speak out loud where you are — a shared flat, a train, a quiet office — typing your answer still gets it marked. It is not the same rehearsal, but it keeps the habit alive on the days speaking aloud is impossible, and a kept habit beats a perfect one you skip.
Trying it
How the speaking marking works explains what it looks at and what it reports back. The drills themselves are at C1 Advanced Speaking practice, part by part, with two free exercises in every part and no account needed.
For the tasks themselves, see the C1 Speaking test explained, the long turn and Parts 3 and 4. The C1 Advanced exam preparation guide covers how much of your final result Speaking actually decides.