WiderAIBlog › How Your IELTS Speaking Score Is Decided, Step by Step

How Your IELTS Speaking Score Is Decided, Step by Step

IELTS Speaking29 August 2026 · 7 min read · WiderAI Team

IELTS Speaking is scored by the certificated examiner who sits opposite you (or appears on your screen), during the test itself, against four equally weighted criteria: Fluency and Coherence, Lexical Resource, Grammatical Range and Accuracy, and Pronunciation. The examiner gives you a whole band from 0 to 9 on each criterion, and your Speaking band is the average of the four, rounded to the nearest half band. There is no computer scoring, no points-per-question system, and no separate mark for each part of the test — the examiner judges your performance as a whole.

That single paragraph answers the search query, but the details matter, because most candidates prepare for a test they imagine rather than the one that is actually marked. Knowing what the examiner is doing minute by minute — and what they are explicitly trained to ignore — changes how you should speak.

Who scores the test, and when?

Your Speaking test is conducted and assessed by one trained, certificated IELTS examiner. They rate you live, forming their judgement during the eleven to fourteen minutes of the interview and finalising the four criterion scores immediately afterwards. The test is also audio-recorded. The recording is not there to be double-marked as routine; it exists for quality monitoring and so that a second, senior examiner can re-mark the test if you later request a review of your result (the official name for this remark process is Enquiry on Results, described at ielts.org).

Examiners are themselves monitored: their scoring is regularly checked against standardised samples to keep ratings consistent between examiners and between test centres. This is why "I got a strict examiner" is rarely the real explanation for a disappointing score — examiners who drift from the standard are retrained. It is also why a friendly examiner is not a generous one: warmth is interview technique, not assessment.

One consequence is worth absorbing early. Because a single examiner rates you live, everything you say from the first "And what shall I call you?" is potential evidence — but only the language counts. Your opinions, your honesty, the interestingness of your life, and whether the examiner personally enjoyed your company have no channel into the score.

What are the four criteria actually measuring?

Each criterion is worth exactly a quarter of your score, so neglecting one costs as much as neglecting any other. In plain English, here is what each one weighs. For a deeper treatment of every band level, see our speaking band descriptors explained guide.

Divide your practice time by criterion, not by test part. Most candidates over-practise Part 2 topics and never once drill pronunciation features or grammatical range — yet each of those silently controls a quarter of the score.

How is the final band calculated?

The examiner awards a whole band (not a half band) for each criterion, guided by descriptors that define performance at every level. Your Speaking score is the arithmetic mean of the four, rounded to the nearest half band. So 7, 7, 6, 6 averages 6.5, and a profile of 7, 7, 7, 6 averages 6.75, which reports as 7.0 under the rounding applied to speaking criterion averages.

Three practical implications follow:

  1. Your weakest criterion drags hardest. A single 5 among 7s pulls the average down a full half band or more, which is why targeted work on your weakest area beats another lap of general practice.
  2. There is no question-level scoring. A weak answer in Part 1 is not a lost point; it is one piece of evidence among fifteen minutes of it. One bad answer never decides the result — a pattern across the whole test does.
  3. Criteria are judged on your best sustained performance, not your worst moment. Examiners expect slips at every band. What they are matching to the descriptors is your typical level across the interview.

What is the examiner doing during each part?

The three parts are not three separate tests; they are three instruments for collecting different evidence against the same four criteria.

Part 1 (4–5 minutes): establishing your baseline

Familiar questions about your life let the examiner hear your natural, unrehearsed level: your pace, your everyday vocabulary, your control of basic tenses. Memorised-sounding answers are counterproductive here — examiners are trained to recognise rehearsed speech and will discount it as evidence, often steering to a fresher question.

Part 2 (3–4 minutes): sustained speech

The two-minute monologue shows whether you can organise ideas and keep going without a partner's support — the purest test of fluency and coherence. The one minute of preparation and the topic card are aids, not part of the assessment; nobody marks your notes.

Part 3 (4–5 minutes): stretching you upward

Abstract discussion questions exist to elicit the language that separates band 6 from band 7 and above: speculation, comparison, concession, hypotheticals. When the examiner pushes with "why?" and "but what about…?", they are deliberately creating chances for complex grammar and precise vocabulary. Give real opinions with real reasons — hedged, extended answers are exactly what this part is designed to reward.

In Part 3, force one complex structure into most answers on purpose: "If governments had invested earlier, cities would be less congested now." One accurate third conditional is worth more as evidence than five safe simple sentences.

What does not affect your score?

Candidates burn energy on things the scoring system cannot see. The examiner does not score your opinions or their political acceptability, your accent as such, the truthfulness of your examples (invent freely — it is a language test), your clothes or body language, small nerves and slips, or whether you needed a question repeated. Asking for repetition in Parts 1 and 3 is cost-free; treating it as forbidden causes the real damage, which is answering a question you did not hear.

What you cannot see, meanwhile, is your own performance profile — which criterion is your ceiling. Guessing it wrong wastes months. An IELTS speaking mock test that scores all four criteria separately shows you exactly where the average is leaking, so your preparation attacks the quarter of the score that is actually holding you back.

After any scored practice test, look only at your lowest criterion and give it 70% of the next fortnight's practice. Raising a 5.5 to 6.5 in one criterion moves your average as much as raising all four by a quarter band.

Practice IELTS Speaking with an AI examiner — full mock test, instant band score and feedback on all four criteria.

Take a free speaking mock test

Frequently asked questions

How is IELTS speaking scored?

A certificated examiner assesses you live against four equally weighted criteria — Fluency and Coherence, Lexical Resource, Grammatical Range and Accuracy, and Pronunciation — awarding a whole band from 0 to 9 for each. Your Speaking band is the average of the four, rounded to the nearest half band. The test is recorded, but the recording is used for monitoring and remarks, not routine double-scoring.

Does each part of the speaking test have its own score?

No. There are no marks per question or per part; the examiner judges your whole eleven-to-fourteen-minute performance against the four criteria. A shaky Part 1 can be outweighed by a strong Part 3, because the examiner rates your sustained typical level, not your worst answer.

Can I ask the examiner to repeat a question, and does it cost marks?

In Parts 1 and 3 you can ask for a question to be repeated, and doing so costs nothing — misunderstanding a question and answering off-topic costs far more. In Part 2 the task is on the card in front of you, so repetition is not needed. Keep requests natural and brief: "Sorry, could you say that again?"

Is it true examiners decide your band in the first minute?

No. Examiners form and revise their judgement continuously across all three parts, and Part 3 exists precisely to test whether your level holds up under abstract, demanding questions. First impressions are real in any conversation, but the score must match the descriptors across the full test, and a candidate who grows into the interview is scored on that stronger evidence too.