IELTS Speaking Part 2: the preparation minute is a planning tool, not a scripting tool
Aman Batth · 10 min read · · Updated
The long turn fails in a predictable way — strong for a minute, thinning at ninety seconds, then a trailing sentence and silence. That is not a confidence problem. It is what happens when you spend the preparation minute writing sentences instead of choosing anchors.
Part 2 gives you a cue card, sixty seconds to prepare, and up to two minutes to speak. Most advice treats those two minutes as a duration to survive.
They are not a duration. They are the amount of connected speech the examiner has to assess you on, and the usual failure is that you hand over less of it than you were offered.
The shape is always the same: strong for the first minute, thinning at ninety seconds, one trailing sentence, silence. Then you wait to be stopped and you are not, because you already stopped.
We build the engine that scores speaking on Hilingo, so this is written from the marking side — not what to say, but what stopping early does to the four criteria. For the descriptors themselves, the source is IELTS, not a blog.
There is no Part 2 score
This is the first thing to fix, because it changes what a bad long turn actually costs.
The examiner does not award a band for Part 1, a band for Part 2 and a band for Part 3. They award one band per criterion for the whole interview, and your Speaking band is the mean of those four, rounded to the nearest half. Our own engine computes it the same way, because that is how the test works.
So Part 2 is not a section you can fail and recover from. It is roughly half of the connected speech in the entire interview — the only stretch where you speak at length without the examiner steering you — and it is the largest single block of evidence feeding all four criteria at once.
A ninety-second long turn does not lose you "Part 2 marks". It thins the evidence behind your whole Speaking band.
What running dry costs, criterion by criterion
The four criteria are equally weighted: fluency and coherence, lexical resource, grammatical range and accuracy, pronunciation. Stopping early does not hit them equally.
| Criterion | What it actually measures | What a short long turn does to it |
|---|---|---|
| Fluency and coherence | Continuity, and whether the answer sequences logically | Direct hit. Less connected speech, and the trailing-off is itself hesitation |
| Grammatical range and accuracy | Variety of structures, not just correctness | Direct hit. Range needs clauses; you never reach the ones that show it |
| Lexical resource | Range within words you control | Little effect either way — see below |
| Pronunciation | Intelligibility across your speech | Little effect; a shorter sample, same sounds |
Two of those deserve explaining, because they are where the advice you have read is wrong.
Fluency and coherence does not reward good ideas
This is the one candidates misread most. Coherence in this criterion means logical sequencing and the use of discourse markers — whether the answer hangs together — not whether the content is interesting or even strictly on topic.
We found this the hard way. An earlier version of our engine blended topical relevance into the fluency signal, and it handed out credit the candidate had not earned: an on-topic answer delivered haltingly scored around a band and a half above the same delivery on a weaker topic. It was measuring the wrong thing, and we removed the relevance term entirely.
The practical consequence: a thoughtful idea delivered in eighty seconds with two long pauses scores below an ordinary idea delivered in a hundred and fifteen continuous, well-sequenced seconds. You do not get fluency marks for having something worth saying. You get them for keeping going and ordering it.
That is why the long turn is the criterion's best opportunity and its worst exposure. Nowhere else in the test do you have two uninterrupted minutes to demonstrate continuity.
Grammatical range needs room to appear
Range is not correctness. A candidate producing clean, short, correct sentences for ninety seconds has demonstrated accuracy and almost no range.
Range shows up as subordination and as variety of construction — relative clauses, conditionals, perfect tenses, passives, comparatives. Our engine measures it directly from the transcript for exactly this reason: how densely you subordinate, and how many distinct constructions you reach for.
Here is the part that matters for Part 2. Those structures do not appear in abstract statements. They appear when you narrate something concrete.
"Travel is beneficial for young people" is a simple sentence and there is nowhere for a subordinate clause to go. "I hadn't planned to stay, but the train I'd booked was cancelled, which is how I ended up spending three days there" contains a past perfect, a relative clause and a contrast — and it came out of a specific detail without any effort to produce grammar.
Abstraction produces simple sentences. Concrete detail produces complex ones. That is the whole mechanism.
Talking longer does not raise lexical resource
Worth saying plainly, because it is the obvious way to game the advice above.
Lexical resource is assessed as range within the vocabulary you actually control. Our engine measures lexical diversity in a length-robust way, so padding the last thirty seconds by recycling the same eight words does not move it. Neither does restating your first point in different order.
Filling time helps fluency and grammatical range. It does nothing for lexis. Which means the fix is not "say more words" — it is "have more things to say", and that is a preparation problem.
So what is the preparation minute for
Sixty seconds. Most candidates try to write sentences. They get four or five words down, run out of time, and then read those four words aloud while their eyes are on the paper rather than on the material.
Two things go wrong at once. The script is far too short to carry two minutes, and delivering the written part sounds different from the unwritten part — which is precisely the kind of discontinuity fluency and coherence is looking at.
Sixty seconds is not enough to script. It is enough to choose. What survives contact with the clock is a sequence of anchors: a rough shape that covers the bullets, with one concrete detail attached to each.
Not "describe the place" but "the yellow kitchen". Not "explain why it mattered" but "my mother's reaction". A noun you can see is worth a paragraph of planned argument, because you can keep talking about a thing and you cannot keep talking about an abstraction.
What to write in sixty seconds
Four or five anchors, in speaking order. One line each. No sentences.
- One anchor per bullet. The bullets are the structure — you do not need to invent one. Give each bullet a concrete noun, name, place or moment you can describe.
- Make at least two of them visual. Something you could draw. Those are the ones that generate description when the analytical material runs out.
- Give the last bullet two anchors, not one. The final line on the card — the "and explain why" — is the one that has to carry the last forty seconds. It is the only part of the card our own engine stores and weights separately from the bullets, because it is where extended speech actually comes from.
- Do not write the opening sentence. You will read it, it will sound read, and the first fifteen seconds are the cheapest part of the answer.
Then put the pen down and look at the examiner.
When you run dry anyway
You will, sometimes. The recovery matters more than the prevention, because trailing off is worse than any of these.
Go back and add detail to something you already said. Not a new point — a specific one. "I said it was a small place; it was actually one room with a window facing the road." This is legitimate, it is what fluent speakers do, and it produces exactly the descriptive clauses grammatical range is measured on.
Contrast it with something. "It is nothing like where I live now, which is …" A comparison is free material and it is structurally complex by nature.
Say what you would change. Hypotheticals give you conditionals, and conditionals are the cheapest range on the card.
Do not announce that you have finished. "So that is all I have to say" hands back thirty seconds of assessable speech and tells the examiner the answer was planned to length. Let them stop you. Being stopped mid-sentence costs nothing.
What this is not
It is not a template. Memorised frameworks — the four-part formula, the fixed opening, the rehearsed connectives — solve the running-dry problem by making every answer the same length and the same shape. They also flatten grammatical range to whatever the template contains, and an examiner recognises the pattern.
The anchors approach is the opposite: the structure comes from the card's own bullets, and the material comes from details only you have.
Hear your own long turn scored
The thing you cannot judge from inside a two-minute answer is whether it held together, because you know what you meant.
Hilingo gives you one full scored mock free, with no card. Speaking is assessed across Parts 1 to 3 the way the test does it — a band on each of the four criteria rather than one number — so you can see whether the long turn is costing you fluency and coherence, grammatical range, or neither. Our engine reads range straight off your transcript, which means it can tell you that you spoke for a hundred and ten seconds in almost entirely simple sentences, and that is a different problem from stopping at ninety.
- IELTS Speaking Part 1 — short answers and why under-producing is the failure there
- IELTS Speaking Part 3 — developing and justifying an abstract position
- The IELTS mistakes you cannot see yourself — why one criterion is usually a band below the rest
- The last week before your IELTS — what actually moves a band late
- IELTS band score chart — band to CLB per skill
Frequently asked questions
How long should I speak in IELTS Speaking Part 2?
Until the examiner stops you. You have up to two minutes and there is no advantage in finishing early — the long turn is the biggest block of connected speech in the whole interview, and a shorter answer simply gives the examiner less to assess. Being cut off mid-sentence is normal and costs nothing.
What happens if I stop talking before two minutes in Part 2?
The examiner moves on. Nothing is deducted as a penalty, but fluency and coherence is assessed on continuity and sequencing, so trailing off and then falling silent is read as hesitation rather than as a finished answer. The cost is real even though there is no explicit deduction.
Can I use my notes while speaking in IELTS Speaking Part 2?
Yes, you are given paper and a pencil and you can look at your notes. The trap is writing sentences rather than notes: four written words read aloud sound different from the rest of your answer, and your eyes end up on the page. Write anchors — a noun or a name per bullet — not prose.
Is one minute enough to prepare for the IELTS cue card?
It is enough to plan and not enough to script, which is the point. Sixty seconds buys you four or five anchors in speaking order, one concrete detail per bullet, with two on the final "and explain" line. It does not buy you a written answer, and trying to write one is why most candidates run out of material.
Can I lie or make up my answer in IELTS Speaking Part 2?
Nothing on the card is fact-checked and truthfulness is not assessed. In practice, invented answers run dry faster, because you have no real detail to fall back on when the planned material is gone — and detail is what keeps a long turn going. Adapting something real is easier than inventing something plausible.
Does Part 2 have its own band score in IELTS Speaking?
No. The examiner awards one band per criterion for the entire interview and your Speaking band is the mean of those four, rounded to the nearest half. Part 2 does not get its own score; it contributes the largest single stretch of evidence to all four criteria at once.