IELTS Speaking: The Complete Guide to Band 7+

If you have taken IELTS more than once and Speaking is the only band still holding you back, you are not a weak English speaker. You are a competent speaker who has not yet been assessed as one. Those are different problems, and they need different solutions.

This guide is for the person who has already banked Listening, Reading and Writing, has sat the test three, four, five times, and keeps landing on 6.5 in Speaking — who has tried a tutor teaching things they already knew, mock tests that never moved the number, and an AI tool that called an answer “great” without hearing the four-second pause in the middle of it.

Below is what is actually measured, why fluent people fail it, and what to do about it. If you want a number first, the free diagnostic at actuallyspeak.com/freeieltstest estimates your band on all four criteria.

Take the free 4-skill diagnostic →

---

Contents

---

How IELTS Speaking is actually scored

Your Speaking band is the average of four separate scores, each worth exactly 25%: Fluency and Coherence, Lexical Resource, Grammatical Range and Accuracy, and Pronunciation. The examiner gives you a whole band on each, and those four numbers are averaged and rounded to the nearest half band.

The arithmetic matters more than almost anything else you will read about IELTS Speaking:

Your four criteria scores	Average	Reported band
7, 7, 7, 7	7.0	7.0
7, 7, 7, 6	6.75	7.0
7, 7, 6, 6	6.5	6.5
7, 6, 6, 6	6.25	6.5

You do not need Band 7 in all four criteria. You need it in three. If you are stuck on 6.5, you almost certainly have two criteria sitting at 6, and the whole job is identifying which two and fixing one. Most candidates never find out: a results slip tells you nothing about where the number came from.

Fluency and Coherence

Not “speaking fast.” This measures whether you can speak at length without effort, and whether what you say hangs together.

Band 6 speaks at length but loses the thread, repeating and self-correcting. Band 7 speaks at length without noticeable effort, and any hesitation left is content-related, not language-related. That distinction is the whole criterion. Pausing to decide what you think is normal, and native speakers do it constantly. Pausing to hunt for a word is a language pause, and an examiner hears the difference. Band 7 also needs discourse markers used flexibly: spoken ones (having said that, the thing is, what I mean is), not written linkers like “moreover” bolted onto speech.

Lexical Resource

This measures range, precision and naturalness, not difficulty. Band 6 has enough vocabulary to discuss any topic at length, but word choice is sometimes imprecise. Band 7 uses vocabulary flexibly, includes some less common and idiomatic language, shows awareness of collocation and style, and paraphrases effectively when the exact word is missing.

Here is the trap that costs experienced candidates most. Inserting impressive words does not raise this score, and using the wrong impressive word lowers it. “Significant”, “substantial” and “considerable” are not interchangeable, and an examiner hearing the wrong one hears imprecision, not range. The reliable Band 7 move is paraphrase: describing around a missing word instead of freezing on it.

Grammatical Range and Accuracy

This measures whether you produce a range of complex structures, and how often they are correct. Band 6 mixes simple and complex forms with limited flexibility. Band 7 uses a range of complex structures with some flexibility and frequently produces error-free sentences, though some mistakes persist.

Two structures do most of the work here, and both are rare in unrehearsed speech unless trained. The first is conditionals (if I had more time, I would…), which Part 3 invites constantly and which untrained candidates replace with flat present simple. The second is flexible comparison (far more… than, nowhere near as… as), because comparison is the backbone of Part 3 reasoning.

Note what the descriptor does not say. It does not say error-free. It says frequently produces error-free sentences. You can make mistakes and score 7.

Pronunciation

The most misunderstood criterion. It does not measure accent. It measures control of pronunciation features and whether you are easy to understand. The features carrying most weight are not individual sounds:

• Word stress on longer words. comFORTable, not comfortABLE. Wrong word stress breaks comprehension faster than most sound errors.
• Sentence stress. Emphasising the word carrying your point, rather than stressing everything equally.
• Chunking. Pausing at grammatical boundaries rather than randomly, which is what makes speech easy to follow in real time.
• Intonation. Rising for uncertainty, falling for certainty. Flat delivery reads as low control even when every word is correct.

You can keep your accent entirely and score Band 8 here. What you cannot do is speak in a flat monotone and score above 6.

---

Why fluent English speakers still get 6.5 in IELTS Speaking

Because fluency in conversation and fluency under assessment are not the same skill, and IELTS scores the second one. You can run work meetings in English, watch films without subtitles, and still produce Band 6 speech in the test, for four reasons.

Your everyday English is efficient, not demonstrative. Real communication rewards getting the point across with least effort. IELTS rewards range. You never need a third conditional at work, so you have stopped producing them, and the examiner cannot award range they do not hear.

You are assessed on your worst two minutes, not your average. In conversation one clumsy sentence disappears. In an eleven-to-fourteen-minute test, one stretch of hesitation in Part 3 is a large share of the whole sample.

Your errors are fossilised. After years of successful communication, a handful of mistakes stopped being corrected because they never stopped you being understood: article use, a tense that drifts, one mispronounced common word. They are invisible to you and audible to an examiner. This is why “just speak more English” stops working at this level — more practice reinforces the same errors.

Nobody has told you which criterion is failing. You get one number, assume the problem is vocabulary, and learn more words. If the real problem was grammatical range or chunking, you have spent three months and another test fee on the wrong thing. That is the pattern behind the most common complaint at this level: a tutor who is pleasant, expensive, and teaching you things you already know.

---

Why you collapse in Part 3 when Parts 1 and 2 went fine

Because Parts 1 and 2 can be survived on preparation and Part 3 cannot. If you prepared by memorising answers, Part 3 is where that strategy runs out. The collapse is not bad luck. It is the design of the test working as intended.

Part 1 is predictable: hometown, work, study, food, weather. Perhaps forty topics, near-identical worldwide, so prepared answers fit. Part 2 gives you a cue card and a minute to plan, and many candidates adapt a rehearsed story to it. Part 3 is a live discussion. The examiner takes your Part 2 theme into abstraction — society, policy, causes, hypotheticals — and builds each follow-up from what you just said. Nothing can be pre-written.

Three things then happen at once, which is why it feels sudden.

The cognitive load doubles. In Part 1 you retrieve a fact. In Part 3 you must form an opinion, justify it, structure it and speak it at the same time. Accuracy that was solid in Part 1 degrades, and pronunciation control flattens out. Holding both steady under Part 3 abstraction is exactly what separates Band 7 from Band 8 in the descriptors.

Memorised material becomes a liability. Examiners are trained to detect memorisation and actively probe for it. A polished, rhythmically odd chunk followed by an unpredictable follow-up is the standard test, and when a fluent stretch is followed by a broken one, the contrast is louder than either part alone. Memorised answers do not just fail to help here. They mark you.

You run out of ideas, not English. Often the freeze is not linguistic at all. Asked whether governments should fund the arts, you have no opinion in any language, so your English carries the thinking load as well as the speaking load, and it buckles. The fix is training idea-fluency — generating a position and two reasons on any abstract topic in seconds — so that when you speak, language is the only job left.

---

What to do when your mind goes blank in the test

Say something out loud immediately. Silence is the only response guaranteed to cost you marks. The recovery move is to fill the gap with real language while your thinking catches up.

Candidates describe this constantly: the mind empties, the pause stretches, and panic makes it worse because now you are also thinking about having stopped. The examiner will not rescue you, and long silence damages Fluency and Coherence directly. Here is the sequence, in order.

1. Buy time with a real phrase, not a filler sound. “Em… em…” is dead air. These are not: That’s an interesting question, I’ve honestly never thought about it. I suppose it depends on what you mean by… They count as discourse markers and buy three or four seconds. Practise them until they are automatic, because under pressure you only produce what is automatic.

2. Narrow the question. Abstract questions freeze people because they are too big. Shrink it out loud: “Well, if I think about my own city specifically…” That is not evasion. It is what fluent speakers do.

3. Take a position, any position. You are scored on the English, not the opinion. Pick the side you have more vocabulary for. If you genuinely have no view: “I don’t have strong feelings about it, but I’d probably say…”

4. Reach for structure, not words. Hunt for a shape rather than a clever word: claim, reason, example. “I think X. Mainly because Y. For example, in my last job…” A structure gives your mouth somewhere to go while your brain restarts.

5. Correct fast, then move on. Fixing a slip in two words and continuing barely costs you. Restarting the whole sentence, or making three attempts at the same clause, is what pulls Fluency down.

Test-day collapse is a trained response, not a personality trait. It happens because the test is the first time you have spoken under real pressure that week. The remedy is exposure: timed, recorded, unpredictable practice.

---

Why your mock scores are higher than your real score

Because most mock tests remove the exact conditions that cost you marks. If you score 7 in practice and 6.5 in the test, the mock is not measuring what the exam measures.

The follow-ups are not real. A genuine Part 3 examiner builds the next question from your last answer. Most mock practice uses a fixed question list, so you are tested on prepared ground, which is the ground you can already handle.

There is no consequence. Your body does not produce a stress response for practice. The version of you at the test centre, after a visa deadline and four failed attempts, is not the version that scored 7 at the kitchen table.

The scorer is lenient, or cannot hear. A friend rounds up because they like you and understood you. AI text tools have a sharper limitation: a tool assessing your transcript cannot assess Fluency and Coherence or Pronunciation at all, and that is 50% of your band. It can call your English Band 7 while you were, in reality, pausing four seconds a sentence.

Nobody scores you criterion by criterion. “That was about a 7” cannot be trained against. Useful feedback names the criterion and the sub-skill: your grammar is Band 7 in Part 1 and Band 6 in Part 3, because accuracy drops when you reason.

Mocks measure. They do not teach. If ten have moved nothing, an eleventh will not help.

---

Is an IELTS Speaking remark (EOR) worth the money?

Sometimes, but only in one specific situation, and you should decide with a clear head rather than out of hope. An Enquiry on Results sends your recorded Speaking test to a senior examiner for a fresh assessment. Your score can rise, stay the same, or in principle fall.

Apply when all of these are true:

• You were within a half band of your target and Speaking was the only component holding you back.
• Your Speaking score was clearly out of line with your other bands. A 6.5 among three 7.5s is an anomaly. A 6.5 among three 7s, again, is a pattern.
• You felt the test genuinely went well. Not “I hoped it did.” If you know you froze, a senior examiner will hear that on the same recording.

Do not apply when you know it was a poor performance, or when the wait will not survive your deadline.

Details vary by centre, so confirm yours: you apply within a set window (commonly six weeks), pay a fee up front, and the fee is refunded if any band changes.

Two honest cautions. Remarks succeed less often than the internet suggests, because people who post about a successful EOR are a self-selecting group. And the money matters: candidates in this position are usually several test fees deep. An EOR is a fair gamble on a genuine anomaly, and a poor use of limited funds when what you need is to raise the performance itself. If you are unsure which it is, the free diagnostic tells you more than a remark will.

---

How to practise IELTS Speaking without a partner

You can train all four criteria alone, and for most candidates at 6.5 solo practice is better than a speaking partner. Partners at your own level cannot hear what an examiner hears, cannot generate real Part 3 follow-ups, and cannot correct errors they share with you. Open partner groups also carry reliability and safety problems. Do this instead.

Record everything and listen back. The highest-value habit, and almost nobody does it, because hearing yourself is uncomfortable. On playback you notice what you cannot notice while speaking: where you paused, how often you restarted a sentence, which word you mispronounced four times.

Count, do not judge. Take one recording and count one thing: pauses over two seconds (content or language?); self-corrections (more than two or three per answer drags Fluency down); complex structures (did a conditional appear at all?); sentence openers (did every sentence start with “I”?). One metric per recording, tracked across a week. That is training. Listening back and thinking “hmm, not great” is not.

Drill the structures you never produce. Answer five hypothetical questions using only conditionals, then five comparison questions using only flexible comparatives. It feels artificial, and that is the point: you are making a structure automatic so it appears later without conscious effort.

Build idea-fluency separately. Take ten abstract Part 3 topics and, without speaking, write two bullets each: your position and one reason. Three minutes daily removes the thinking load from your speaking.

Shadow for pronunciation. Take thirty seconds of a native speaker. Play a sentence, pause, repeat it copying the rhythm and stress rather than the individual sounds. This trains stress, chunking and intonation directly, which is where the marks sit.

Simulate pressure. Timer on, no second attempts, no looking words up. One take, exactly as in the test.

What you cannot do alone is find your own fossilised errors, invisible to you by definition, or generate unpredictable follow-ups. That narrow job needs a trained ear.

---

What separates Band 6.5 from Band 7, and how long it takes to close

The gap at this level is never “your English is not good enough.” It is three or four named sub-skills, you probably need only two, and for someone already speaking English adequately the realistic window is eight to twelve weeks — but only if the work is targeted. Repeating general practice can hold you at 6.5 for years, which is why some people sit the test seven times.

The sub-skills are the ones named under each criterion above: spoken discourse markers, conditionals without hesitation, sentence stress, chunking, precise word choice at speed. None of that is “learn more English.”

The training sequence is always the same, and the order is not optional: diagnose which two criteria sit at 6, name the sub-skills, drill each in isolation until it is automatic, then re-test under pressure — because a skill that only appears when you concentrate on it will not appear on test day.

Step one is free: take the diagnostic. The rest is what our IELTS coaching programme is built to do. Speaking-first by design, one-to-one plus small group, and deliberately not grammar drilling, because that is what most people at 6.5 have already tried. Programmes run over 12, 24 or 36 weeks depending on your deadline.

Matthew Fabling has coached English since 2011 and worked with more than 500 students across China, Korea, the UK, Mexico, Brazil and Europe. The consistent finding: candidates stuck below target rarely need more English. They need someone to hear the specific thing they cannot hear themselves.

---

Frequently asked questions

---

The next step