Short answer: the Speaking Part 2 question bank rotated on 1 September 2026, and the set below is what candidates are reporting for the September–December window.
Most rotations keep a large share of the previous set, so topics from May–August are still worth preparing.
A forecast tells you what to rehearse; it does not tell you what you will be asked.
How this forecast is made — and what it is not
Let us be exact about this, because the category is full of pages that are not. IELTS does not publish its question bank, and nobody has access to live exam content.
What exists is candidate recall: people who have just sat the test post the cue card they were given to forums, comment threads and teacher blogs, and aggregators collect those reports.
This page collates what those public sources are reporting for the current window, cross-checks a topic against more than one source before listing it, and links each group to material you can actually work with.
Three honest limits follow from that method. Reports are self-selected, so common cards are over-represented and rare ones may be missing entirely.
Question banks vary by country and by test centre, so a topic reported heavily in one market may not appear in yours.
And the bank is large — knowing the pool is not the same as knowing your card.
Treat this as a rehearsal list, not a prediction. Anyone promising you the questions is either guessing or lying.
The useful part is that the pool is genuinely finite and genuinely repetitive.
Cards recur across rotations with the nouns swapped — a person, a place, an object, an occasion — and a candidate who has rehearsed one card per category can adapt to almost anything the examiner produces.
What is being reported for September–December 2026
Grouped by the underlying task, because that is what actually transfers. The right-hand column is the nearest card in our library — not the identical wording, but the same job.
| Reported prompt | What it really tests | Practise with |
|---|---|---|
| A place you would recommend others live in (not your hometown) | Describing a place + justifying a recommendation | A city you would like to live in |
| Someone else's room you enjoy spending time in | Describing a space and why it feels the way it does | Your favourite room |
| The home of someone you know well and often visit | Same skill, different owner — description plus relationship | A house or apartment you would like to live in |
| A person others like to ask for advice | Describing character through behaviour, not adjectives | An intelligent person you know |
| Something in your study or work that made you feel confident | Narrating an experience with a feeling attached | An achievement you are proud of |
| A place in your city where you can see wildlife | Place description with a specific, concrete focus | A park or garden you like |
| An occasion when you were not allowed to use your mobile phone | Narrating a constrained situation and your reaction | Something you use every day |
| An item you spent more on than you expected | Object description plus a decision you can justify | Something you bought recently |
Two patterns are worth noticing. First, several of these are places seen through somebody else — a room that is not yours, a home you visit, a person others consult.
That shift catches candidates who have only rehearsed first-person favourites, because the vocabulary is the same but the framing is not.
Second, the constraint prompts ("not your hometown", "not allowed to use your phone") are doing deliberate work: they block the pre-memorised answer.
An examiner who hears a rehearsed speech that ignores the constraint marks it down under Task Response, however fluent it is.
How to use a forecast without wasting your preparation
The mistake is memorising.
A candidate who learns eight answers by heart has eight answers and no flexibility, and examiners identify recited material quickly — the rhythm flattens, the vocabulary jumps a level above the rest of the interview, and the answer stops matching the card the moment a constraint is added.
What works instead is preparing material rather than scripts.
Take five or six experiences you can describe in real detail — a trip, a person, a purchase, a place you return to, something you made or learned — and rehearse each one until you can attach it to different prompts.
The trip becomes a place, a memorable day, a decision you made, a time you were surprised.
That is the whole trick, and it is why our cue cards carry an idea map as well as a model answer: the model shows you the register, and the map shows you the angles you could take instead.
Then rehearse against a clock. One minute to prepare, one to two minutes to speak, no restarts.
Our timed cue-card practice runs exactly that shape, and grades what you say against the four official criteria — Fluency and Coherence, Lexical Resource, Grammatical Range and Accuracy, and Pronunciation (estimated) — so you find out which one is holding your band down rather than guessing.
Parts 1 and 3 rotate too
Almost every forecast page you will find covers Part 2 only, because a cue card is easy to recall and repeat.
Parts 1 and 3 rotate on the same schedule and matter just as much: Part 1 sets the examiner's first impression, and Part 3 is where the band is usually won or lost, because it asks for abstract, extended reasoning rather than personal description.
Each of our cue cards therefore carries its Part 1 warm-ups and its Part 3 follow-ups with model answers on the same page — the topic threads through all three parts in a real test, and it should thread through your practice the same way.
When this page changes
The next rotation is January 2027. We refresh this page as reports accumulate through the current window and replace it when the bank turns over.
If you are sitting the test in the next few weeks, work the groups above; if your test is in January, prepare the categories rather than the cards, because the specific prompts will move and the categories will not.