Speaking — The Complete Guide

Everything the Speaking test needs, in one place: the four criteria, the shape and length of an answer in each part, the phrase banks, the recovery lines, the emergency protocols, and the myths that waste preparation time.

The one idea underneath all of it: Speaking is not a knowledge test, it is a performance test. Every section below is about producing controlled English under time pressure, not about having better opinions.

Fully worked model answers live in the part-specific pages linked at the end.

Sources. Original material, plus condensed and adapted guidance from IELTS Advantage (Christopher Pell) and IELTS Podcast (Ben Worthington). The myths table follows IDP and British Council published guidance; the criteria wording is the official version.

1. What the test actually is

One continuous 11–14 minute block. It is recorded. The examiner has a script for Parts 1 and 2 and freedom in Part 3.

2. The four criteria — 25% each

Fluency and Coherence

Fluency means speaking smoothly at a natural pace — not fast. Native speakers pause often; that is normal. Two failure modes:

• Speaking too fast. Your brain cannot keep up with your mouth in a second language, and coherence collapses. Copy the pace of radio and film, not a faster one.
• Over-focusing on grammar mid-sentence. This produces the hesitation and self-correction that Band 5 responses are full of.

A test worth running. Record yourself answering one question focused on accuracy, then a similar question focused purely on flow. Most people are surprised that the "sloppier" version sounds clearer overall.

Coherence means answering fully and linking points logically. "What causes pollution here?" — Band 5 says "traffic and industry." Band 8 explains why each causes pollution and gives a concrete instance.

Discourse markers work like cohesive devices in Writing: Band 5 overuses or misuses them (opening with "in my opinion" and then not giving an opinion); Band 7–8 uses them only when needed.

Lexical Resource

Band 5 describes vaguely — "some different aspects." Band 8 uses topic-specific words accurately and rarely repeats itself.

The 100% rule. Only use a word in the exam if you know its meaning, its grammar, its collocations and its pronunciation completely. A misused rare word signals a limited vocabulary, not an advanced one — it is one of the clearest tells of a weak lexicon.

Grammatical Range and Accuracy

A mix of simple and complex sentences, but accuracy matters more than ambition. Band 8 uses a wide range with very few errors; Band 5 collapses into errors the moment a complex sentence starts.

Identify your own recurring mistakes rather than trying to fix "grammar" as a whole.

Pronunciation

You are not marked on your accent. Accent is where you are from; pronunciation is whether the sounds land. The criteria never mention accent, and faking a British or American one wastes preparation time.

What is actually assessed:

• Word stress — PHOtograph vs. phoTOGrapher. Getting it wrong makes a correct word unrecognisable.
• Sentence stress — content words (nouns, verbs, adjectives) get stressed; function words don't.
• Weak sounds — unstressed function words shrink toward /ə/. This is what makes speech sound smooth rather than robotic.
• Linking — words run together at their boundaries: "bit of a" → "bitova".
• Intonation — flat intonation reads as disengaged even when the words are correct. Pitch rises mid-sentence and falls at the end.

Micro-listening. Find a native-speaker recording, pause on any phrase you didn't catch, and replay it until you can transcribe exactly what was said and identify the linking and weak sounds that obscured it. It trains listening and pronunciation at the same time.

At Band 9 the examiner is listening for three layers working together, not accurate sounds in isolation: individual sounds distinct enough that ship/sheep and van/fan don't collapse into each other; the stress and rhythm above them; and intonation carrying meaning across the whole utterance. Training with minimal pairs builds the first layer; shadowing a recording builds the other two.

2b. Before you speak a word

First impressions set the tone, and a few small politeness habits read as fluency:

• Match the greeting to the time of day — "good afternoon".
• Wait to be offered a seat. If you must ask, the polite form is "Please may I sit" — can signals authority rather than permission.
• Answer your name with a contraction: "My name's…", not the stiffer "My name is…".
• Speak slowly. Speed reads as nervousness and makes you harder to follow — slowing down also buys you thinking time.
• Let the examiner finish the question before you answer.
• Don't ask the examiner questions back ("and how are you?"). That is not your role in this exchange.

3. Part 1 — the target shape

Answer. Reason. One concrete detail. Stop.

30–45 words. About two breaths. Never a third reason.

Too short — 15 words:

Yes, I do. It is interesting work and I have been doing it for years.

Correct — 38 words:

Yes, very much so. It is demanding work, but no two days look the same, which suits me. I would probably get restless in something more predictable, so the unpredictability is the part I would not trade.

Too long — 70 words:

…so the unpredictability is the part I would not trade. Also, I work with people at a difficult point in their lives, and that changes how you see almost everything else. When I started, I found that part harder than I expected, although now I would say it is the reason I stayed.

Everything after the first version was unnecessary — the answer was already complete. And notice what happened in the extra sentences: the tense moved to the past, then the present perfect, then a conditional, all inside sentences nobody asked for.

That is where tense control is lost. Not at the start of an answer — in the fourth sentence.

The drift alarm — three signals

1. You are starting a third sentence and it begins with and, also, or another thing.
2. You have started explaining background nobody asked for — when you were younger, how it started, what it used to be like. That is Part 2 material.
3. You have qualified your own qualification — although, having said that, then again, all in one answer.

The landing lines

If the alarm goes off mid-sentence, don't stop dead. Land it:

So yes, on the whole I would say I do. · So that is really the main reason. · So it works well enough for me. · So I would say yes, broadly. · So that is where I stand on it.

A landing line converts a drift into a conclusion. Nobody hears the difference.

Opening patterns by question type

Matching your opening to the question type signals in the first few words that you understood it — and sets the right grammar. Adapt these; don't recite them.

4. Part 2 — the two minutes

At a comfortable pace, two minutes is roughly 280–300 words. Fewer than you think.

Across the spine, that is roughly 120 words of past, 100 of present, 60 of future and reflection — because a top answer has to demonstrate tense range, and range is what a past-only story cannot show.

Two checkpoints you should feel without looking

Aim to be stopped. A good Part 2 answer ends mid-sentence because the examiner cuts in.

The preparation minute

Four scribbles, not sentences. Four fragments generate 300 words, because the spine does the rest:

:00–:10   Name the topic + your direct answer.
:10–:45   The story or event — past tense, one time, one place, specific detail.
:45–1:30  Expand: how, why, what happened next.
1:30–2:00 Reflect: what it means now, and what it means going forward.

The bullet points on the card are prompts, not a checklist. Speaking fluently for the full time on the topic matters more than ticking every box.

The two ways Part 2 fails

Running dry — 60 words, then silence:

I would like to talk about a doctor I worked with during my training. He was very strict but he taught me a lot. He used to ask a lot of questions and everybody found it difficult. I learned a lot from him. He was a good teacher.

No anchor, no scene, no single event, no present, no future. The repair is not more effort — it is one concrete event with a time and a place. One event generates a whole minute by itself.

Drifting — 120 words and no story has started:

…although actually to explain properly I should say something about how the training system works here, because it is quite different from other countries. You do several years and you rotate through different services, and there were staffing problems that year as well…

At 45 seconds this candidate is still explaining a system. They will be stopped in the background section, having demonstrated one tense.

The repair is checkpoint one. If you notice yourself explaining a system, an institution, or how something generally works, you have left the card.

The recovery lines

Anyway, the reason I mention it is… · But the particular occasion I am thinking of was… · To give you a concrete example of that… · What actually happened was…

Each drops you straight back into a specific past event with a specific verb. That is all a recovery needs to do.

5. Part 3 — the discussion

A discussion, not a monologue. Four to six exchanges over 4–5 minutes. You are not trying to fill a fixed block of time and you are not trying to be stopped — you finish your answer and hand the turn back.

The questions escalate deliberately. The first set asks you to describe or compare (targeting Band 6). The second asks you to analyse or explain causes (Band 7). The third asks you to hypothesise or discuss something abstract (Band 8–9).

This means something reassuring. If the questions turn strange or difficult near the end, that is not the examiner losing patience — that is the examiner checking whether your ceiling is higher than your previous answers proved. Difficult questions late in Part 3 are a compliment.

The target

Four to five sentences. Roughly 100–130 words. About 45–55 seconds.

Under three sentences, you have not shown analytical range and the examiner moves on early, which costs you turns. Over about seven, three things go wrong at once: you start a second argument you cannot finish, the tense drifts because you have left the frame the question set, and you burn exam time in which you could have shown range across more questions. A long answer is not a generous answer.

The spine — assertion, mechanism, evidence, concession

Four moves, in this order, every time, whatever the question.

1. Assertion. Answer in one sentence, immediately, with a position in it. No preamble, no repeating the question back.
2. Mechanism. Explain why or how. This is the move that separates Band 7 from Band 6, because it shows you can explain a process rather than only state a preference.
3. Evidence. One generalised, real-world illustration — a country, an industry, a demographic, an institution. Never a person you know.
4. Concession. Name the limit of what you just said. This does two jobs: it demonstrates analytical range, and it ends the answer for you.

The concession is your exit. Once you have conceded, you stop. That is the whole discipline, and it is the structural fix for drift.

One spine, not five formulas. Under pressure you cannot select between competing templates, and the selection burns the seconds you needed for content. AWEA is the same four moves under different letters. REE is this shape with the concession removed, which is why it produces shorter and weaker answers. CTME is a quality checklist, not a structure. Multiple Perspectives is a content generator, not a structure.

The six question archetypes

The spine doesn't change. What changes is what goes into moves two and three.

Where ideas come from — three lenses

When the topic is unfamiliar, you don't need an opinion about it. You need a lens.

Lens 1 — the three perspectives. Look at the question as the individual, the economy, and society or the environment. Almost any topic yields something under all three. Private cars are convenient for the individual, valuable to a manufacturing economy, and damaging environmentally — that single move generates a mechanism, an example and a concession in one pass.

Lens 2 — institutions vs. individuals. Ask who has the power here and who bears the cost. Governments, employers, schools and hospitals on one side; households, workers, students and patients on the other. Any question about a social problem can be answered by describing how those two levels interact, and it produces exactly the vocabulary examiners are listening for.

Lens 3 — the professional generalisation. Whatever you actually know well — a trade, an industry, an institution you have worked inside — is usable, but only if you strip the first person from it. Not "when I write evaluations for my trainees." Instead: "when supervisors design evaluation criteria for trainees, the difficulty is usually that…" Same expertise, correct register, and more specific than anything a general candidate can produce. The only barrier is the pronoun.

On using "I"

The rule that says "never use I in Part 3" is overstated, and as usually written it is dangerous.

The real rule is narrower: do not use personal anecdote as evidence. Your cousin, your commute and your weekend are not data. But the examiner is explicitly asking for your view, and refusing to give one sounds evasive rather than academic.

Say I think, I would argue, my sense is — then support it with something societal rather than something autobiographical. The target is a personal opinion supported by impersonal evidence.

Handling pushback

The examiner will sometimes challenge you: "But some people would say the opposite." / "Do you think that is always true?"

This is a probe, not a disagreement, and it is a good sign. Do not abandon your position and do not defend it rigidly. Concede the specific point, hold the general one.

That is fair, and in the case of X it probably does hold. What I would say though is that in most cases the stronger factor is still Y.

Three phrases: That is fair, though I would still argue… · There is definitely something in that, but… · I would agree up to a point, and the point where it stops holding is probably…

Emergency protocols

1. Honest ignorance with an attempt — for a strange or unfamiliar topic.

To be honest, this is not something I have thought about much. But if I had to guess, I would say… → then answer for three sentences.

The examiner is assessing whether you can navigate the unknown in English. Admitting ignorance fluently while still producing an argument is a strong performance, not a weak one.

2. The conditional answer — when a yes/no fits neither way.

It really depends on. If the situation is this, then this follows. Whereas if it is that, the outcome is different.

That is a full answer with an assertion, two mechanisms and a built-in concession.

3. Geographical narrowing — when the question is too broad to grip.

I cannot speak for everywhere, but at least in my country, what has changed most is…

Narrowing is not evasion. It is what a careful speaker does with an unanswerably broad question.

4. The definitional reframe — for abstract questions. The single most effective move against a Band 9 probe.

That depends what we mean by success. If we take it to mean financial security, then… If we mean something closer to autonomy, the answer is different.

6. The phrase bank

Short on purpose. Ten phrases you can actually deploy are worth more than fifty you have to remember to insert.

Phrases to avoid

These are written register. Spoken aloud in a conversation they sound recited — and examiners are trained to detect exactly that:

I am of the firm conviction that · It is abundantly clear to me that · There is compelling evidence to suggest that · A critical analysis of this trend reveals that · A clear manifestation of this is · To illustrate this point, we need look no further than · Without a shadow of a doubt · A textbook example of this phenomenon.

Also avoid firstly, secondly, in conclusion — essay connectors. Say the main reason is, another thing is, so on the whole.

Thinking-time phrases that sound natural, not hesitant

If I remember correctly, it was… · Let me think about how best to put this… · I guess you could say… · That's an angle I haven't really considered before, but…

Idioms — a warning, not a requirement

Idioms are the most over-hyped feature in Speaking preparation. "Idiomatic" in the rubric means natural — ordinary everyday phrasing, phrasal verbs included — not literally raining cats and dogs. Seven ways they actively hurt a score:

1. Using them in Writing Task 1 or a formal letter — always the wrong register.
2. Using them in Writing Task 2 — same issue; Task 2 is academic prose.
3. Overusing them in Speaking. Native speakers use idioms far less often than students think. Stacking several into one answer sounds unnatural and starts costing you coherence.
4. Using one you don't fully understand — saying "I go there once in a blue moon" and then describing doing it weekly tells the examiner you don't know the phrase.
5. Saying it wrong — "hit the nail with the hammer", "raining dogs and cats". Worse than not attempting one.
6. Thinking about idioms mid-answer. The mental effort of searching for one costs you fluency.
7. Believing they raise your score on their own. They do not.

The 100% rule applies here too: only use an idiom in the exam if you are completely sure of its meaning, its exact wording, and its context.

A few that are genuinely common, used sparingly:

Fixed collocations are the safer bet. Bear in mind, take into account, as far as I'm concerned — they sound native, they reduce your error rate, and they carry none of the risk above.

Asking for clarification

Fine: ask the examiner to repeat a question, or to explain the meaning of one word. Not fine: ask them to explain a whole sentence, ask them personal questions back, ask them to change the question, or use it as a stalling tactic every time.

Sorry, could you repeat the question, please? · Would you mind explaining what ___ means?

If you need thinking time, a three-step move works well: paraphrase the question back (which shows off a real skill), briefly admit you need a moment, then attempt an answer.

That's a tricky one to answer straight away… give me a second to think… okay, I suppose I'd say…

7. Grammar targets

Three things to make audible, chosen because this is where marks are actually available.

1. The cleft sentence. It is the media that shapes this rather than the media shapes this. It is the cost of housing that drives the decision. One or two per Part 3, no more, and each carries a visible Band 8 marker. It is the same structure you drill in Task 2, so it costs you nothing new.

2. Agreement with abstract singular subjects. Part 3 runs almost entirely on abstract nouns, which is why the -s disappears here more than anywhere else:

Access to healthcare determines… · Migration drives… · The cost of housing shapes… · Automation replaces… · Technology changes…

Every one is singular with a plural noun sitting nearby.

3. Tense control in comparison questions. When the question compares past with present, you will start in the past and stay there. The fix is structural, not mental: put the past in the mechanism sentence and the present in the evidence sentence, and let the paragraph shape move you rather than trying to remember mid-sentence.

Self-correction

The descriptors reward self-correction that resolves cleanly. What lowers your fluency score is looping — repairing the repair, or apologising for the slip.

Repair once, briefly, and continue. "Sorry — took, not take." Then keep going. Never repair the same sentence twice, and never say sorry twice.

8. Seven myths that waste preparation time

Per IDP and British Council's own published guidance, these widely repeated tips are false or actively harmful, because they redirect study time away from the four criteria.

The pattern behind all seven: clickbait tips spread faster than accurate advice and get copied without verification. When in doubt, the only source that matters is the marking criteria.

Seven fluency problems and their fixes

1. Speaking too fast → slow to a natural pace. It buys thinking time too.
2. Worrying about having "good" ideas → there is no intelligence test here. Any relevant, honest answer works.
3. No structure for Part 2 → use the spine in §4. The bullets are not a limit.
4. Over-focusing on grammar mid-answer → record two answers, one for accuracy and one for flow, and compare which sounds better overall.
5. Not enough practice → the skill improves with use, not passive study.
6. Trying to eliminate "umm" entirely → these are natural in every language. Only excessive hesitation is a problem.
7. Nerves → mostly a symptom of under-preparation. The fix is the same as for 1–6.

9. The closing checklist

Part 1 — 30–45 words. Answer, reason, one detail, stop. Never a third reason.

Part 2 — 280–300 words. Story finished by 0:45. Present marker by 1:30. Aim to be stopped.

Part 3 — 100–130 words. Assertion, mechanism, evidence, concession. The concession is the exit.

Throughout — the -s on third-person singular verbs, especially after abstract nouns. Repair once, never twice. A clean simple sentence outscores a broken complex one.

Where to go next