Book the computer-delivered IELTS and you will type your essays, click your Reading answers — and then walk into a room and talk to a person. The official format page is blunt about it:
"The Speaking test is a face-to-face interview between the test taker and an examiner. The Speaking test is recorded."
IDP says the same thing about the computer test specifically — "The IELTS Speaking remains the same, with a face-to-face interview with an examiner." The one variant is IELTS Online, where ielts.org says "Your Speaking test will still be with a human examiner, and happens via an online video call" — a live video call, not the in-person session, but still a person. In no delivery mode does a machine assess IELTS Speaking. And the whole interview is recorded — that recording is the permanent record of your performance, which is exactly why "the examiner seemed distracted" is never the end of the story.
The failure
The mistake this page prevents has a name: rehearsing a performance for a conversation. Candidates script answers to predicted questions, memorise them, and deliver them at a wall — which works on a machine-scored test and fails in front of a trained human. IDP's official line:
"The examiner will be able to tell if you have memorised your answers and this may influence your final band score."
The same IDP article adds that "it is never advisable to memorise answers in preparation for your IELTS test" and that examiners "are well trained to spot them." That is the full extent of the official wording — no one publishes the exact penalty — but it is enough. The test is built as an interview, in three parts, so that scripts run out.
The three parts
| Part | Official time | What happens |
|---|---|---|
| 1 — Interview | 4–5 minutes | The examiner checks your identity, then asks "general questions on some familiar topics, such as home, family, work, studies or interests" |
| 2 — Long turn | 3–4 minutes, including the preparation time | A task card, one minute to prepare with pencil and paper, then you "keep talking for 2 minutes" until the examiner stops you |
| 3 — Discussion | 4–5 minutes | You "discuss issues related to the topic in Part 2 in a more general and abstract way and, where appropriate, in greater depth" |
All wording above is from the official format pages. The whole test runs 11–14 minutes — the official IELTS Guide for teachers gives that total. Three details worth noticing in the fine print:
- Part 2's "3–4 minutes" already includes your one minute of preparation. The talk itself is "up to two minutes" — the examiner "will stop you when the time is up", so being cut off is the timing working, not a bad sign. "They may then ask you one or two questions" before Part 3.
- Part 1 starts with an identity check. The first minute is admin, not assessment theatre — the familiar-topics questions follow.
- Part 3 is chained to Part 2. The discussion grows out of your task-card topic, moved to general and abstract ground — which is why Part 3 cannot be predicted from question banks.
Four criteria, equally weighted, across all three parts
You are not scored part by part. The official scoring page names four criteria — Fluency and coherence, Lexical resource, Grammatical range and accuracy, Pronunciation — and states:
"Each of the criteria carry equal weighting and the overall average gives the IELTS score for Speaking."
25% each, averaged. There is no separate Part 2 mark to protect or Part 1 mark to coast on — the examiner rates the whole 11–14 minutes against the same four scales.
On accent, quote the descriptors, not the folklore. IELTS publishes no blanket "accent is not penalised" statement — we could not find one anywhere on the official domains. What the current official Speaking descriptors actually say, at Band 8 Pronunciation, is:
"Accent has minimal effect on intelligibility."
Accent enters the scale only through intelligibility — through whether you can "be easily understood throughout." IDP adds, of the face-to-face format, that "the IELTS examiner understands a wide range of accents so will be able to understand what you say, unlike an AI machine." So the honest version is narrower than the folklore: your accent is not the target; being understood is.
What carries across all three parts
Our coaching, not official wording — but each point leans on an official line:
- Answer the question asked. Part 1 asks about your home, work and interests; Part 3 asks for general, abstract discussion. A Part 3 answer that retreats to personal anecdote is answering a Part 1 question nobody asked.
- Keep speaking. The Band 6 Fluency and coherence descriptor rewards being "able to keep going and demonstrates a willingness to produce long turns." Silence gives the examiner nothing to rate on any of the four criteria.
- Self-correction is normal speech, not a confession. Band 7 Fluency and coherence says it outright: "Some hesitation, repetition and/or self-correction may occur… However, these will not affect coherence." Fixing a sentence mid-flight and moving on is described inside Band 7 — do not let one stumble collapse the turn after it.
What is not official
- There is no official raw-score or conversion table for Speaking. Listening and Reading have official "average marks" tables; Speaking is examiner-rated straight onto the four band scales. Any "Speaking score calculator" you meet elsewhere is not official.
- Two official descriptor versions circulate. An older Cambridge-hosted "public version" words the pronunciation scale differently ("L1 accent"). This page quotes only the current ielts.org PDF — if a guide mixes the two, treat its quotes with suspicion.
- The recording's exact role in disputes is not spelled out in our sources. The official pages confirm the test "is recorded", full stop. If you are considering disputing a score, check the official Enquiry on Results information with your test centre before paying anything.
What actually costs you
- Delivering a memorised speech to a trained human — official IDP wording says it "may influence your final band score".
- Short answers in Part 1 that never show a long turn — "willingness to produce long turns" is in the descriptors from Band 6.
- Treating being stopped in Part 2 as failure and deflating for Part 3 — the stop is the format.
- Chasing an accent instead of intelligibility — the scale only asks that you be understood.
- Freezing after a self-correction — the descriptors treat self-correction as ordinary speech.
Try one
The fastest way to feel the three-part rhythm is to sit one timed Speaking task and hear yourself against the four criteria.
[Try a Speaking task →] — one full task, marked, no card required.