Search for IELTS Writing Task 2 topics for 2026 and you will find lists of predicted questions, some of them sold to you. We are not going to publish one. Nobody outside the exam board knows what is on your paper, and anyone who says otherwise is guessing with your money.
There is a better reason to skip those lists. The topic is the easy half. Task 2 recycles a small set of subjects year after year, and you can build a working stock of ideas for all of them in about a week. What actually decides your Task Response band is the question type: the shape the instruction sentence asks for. Two candidates can draw the same subject, write English of the same quality, and finish a band apart because one of them built the wrong essay.
So this guide does two things. It gives you the topic families that keep coming back, taken from our own Task 2 prompt bank rather than from a prediction. Then it shows you the six question shapes that cut across every one of those families, with two worked examples where a single topic turns into two completely different essays because the instruction changed.
#What we counted, and what it does not mean
Our writing bank holds 287 prompts. 150 of them are Task 2 essays: 90 Academic and 60 General Training. We sorted all 150 twice. Once by subject, and once by the instruction sentence that closes the prompt.
Be clear about what that is and is not. These are prompts we wrote to imitate the real paper. They are not leaked questions, they are not a survey of real exams, and no count below predicts your test. What a bank like this does show is which subjects a serious Task 2 syllabus has to cover, and how unevenly the question types are spread across them. The second finding is the one that should change how you prepare.
#The eight topic families
Sorted by subject, the 150 prompts fall into eight families. The counts are our bank's, not the exam's.
- Education, 31 prompts. Tuition fees, compulsory subjects, single sex schooling, university rankings, whether universities should teach job skills or theory.
- Work and money, 27 prompts. Working from home, the four day week, retirement age, salaries, advertising, consumer behaviour, automation.
- Environment, 22 prompts. Plastic waste, renewable energy timelines, individual versus government responsibility, conservation, animal welfare.
- Technology, 20 prompts. Social media, artificial intelligence, screen time, online learning, whether platforms are responsible for what users post.
- Society and government, 16 prompts. Ageing populations, urban housing costs, public spending choices, compulsory voting, family life.
- Health, 15 prompts. Food advertising and obesity, sedentary lifestyles, mental health provision, public versus private healthcare.
- Globalisation and culture, 9 prompts. Tourism, migration, museums and the arts, local versus world history.
- Crime and law, 6 prompts. Prison sentences versus rehabilitation, whether education reduces crime.
Two things follow. First, the tail is short. Eight families cover it, so a small stock of ideas, examples and vocabulary for each one means you will never meet a prompt you have nothing to say about. That is the real benefit of topic preparation and it is worth a few evenings.
Second, look at how small the crime family is next to education. Candidates who spend a month collecting prison arguments are polishing a corner of the syllabus. The families that keep coming back are the ones closest to ordinary adult life: school, work, money, the environment, and the phone in your hand.
#Prepare arguments, not essays
The useful unit of preparation is not a memorised essay. It is a transferable argument move. Four of them cover a surprising amount of our bank:
- Cost and opportunity cost. Public money spent here is not spent there. Works for free university, public art, space programmes, loss making bus routes.
- Targeting beats blanket policy. Help the people who need it rather than everyone. Works for tuition fees, subsidies, welfare, healthcare.
- Behaviour follows incentives, not instructions. People change when the cheap option changes, not when they are told to change. Works for car use, recycling, diet, remote work.
- Short term cost against long term benefit. The bill lands now and the payoff lands in twenty years, or the reverse. Works for climate policy, education, pensions, research funding.
Each move is a paragraph skeleton you fill with whatever topic arrives. That is preparation that survives a surprise. A memorised essay is not, and examiners recognise memorised material instantly.
#The six question types, and why they matter more
Now sort the same 150 prompts by their instruction sentence instead of their subject. This is where the distribution stops being even:
- Discussion, 61 prompts. "Discuss both views and give your own opinion."
- Opinion, 42 prompts. "To what extent do you agree or disagree?"
- Advantage and disadvantage, 17 prompts. "Do the advantages outweigh the disadvantages?"
- Two part, 16 prompts. Two separate questions, neither of them an agree or disagree.
- Problem and solution, 12 prompts. "What are the causes, and what can be done about them?"
- Cause and effect, 2 prompts. "What are the causes, and what effects does it have?"
Discussion and opinion together are 103 of the 150. If your practice time is short, that is where it goes, and our opinion essay structure guide takes that shape apart paragraph by paragraph. But do not read the small numbers as permission to skip the rest. A problem and solution prompt you have never planned is the one that eats fifteen of your forty minutes while you decide what the essay is supposed to look like.
#The instruction is often not the textbook line
Here is the trap inside the trap. Only 93 of our 150 prompts carry one of those standard instruction lines word for word. The other 57 ask for exactly the same shapes in different words. These are real closing lines from the bank:
"Do you think voting should be compulsory?"
"Who benefits more from this arrangement, employees or employers?"
"Which group do you think gains more from working from home?"
"What is your view?"
All four are opinion essays, and not one of them says "to what extent do you agree or disagree". A candidate who has memorised five labels and nothing underneath them will default to discussing both views, and hand in a balanced answer to a question that asked them to choose. That costs Task Response marks before the first sentence is written.
#Worked example 1: working from home, two shapes
Our bank holds eight Task 2 prompts about working from home. Take two of them.
Prompt A. "Some companies now allow their employees to work from home permanently, with no requirement to attend a physical office. Do the advantages of this trend outweigh the disadvantages?"
Prompt B. "In recent years, working from home has become much more common in many countries. Why has working from home become more common? Do you think this trend will continue in the future?"
Same subject. Same ideas available to you. Two essays that share almost nothing structurally.
#Prompt A is a weighing task
"Outweigh" is a word about balance. The examiner wants a verdict on that balance, held from the introduction through to the conclusion. Advantages and disadvantages sitting politely side by side, closing with "in conclusion, there are both advantages and disadvantages", is the classic band 6 answer to this type. The public band descriptor for Task Response at band 6 says the response "Presents a relevant position although conclusions may become unclear or repetitive." A list produces exactly that.
The plan:
- Introduction. Paraphrase, then state the verdict, with one condition attached.
- Body 1. The strongest advantage, developed: time and money returned by not commuting, and what people do with it.
- Body 2. The real disadvantage, conceded honestly, then weighed and overruled.
- Conclusion. The verdict again in different words, plus the condition.
Here is body 2 written out, because conceding and then weighing is the move most candidates cannot make:
"The most serious cost of permanent home working falls on new employees. Junior staff learn a great deal by accident. They overhear how a colleague handles a difficult client, and they ask small questions the moment those questions occur to them. Video meetings remove almost all of that, because nothing is overheard and every question has to be scheduled. This is a genuine loss and it is the strongest argument against the trend. It is not, however, a loss that outweighs the benefits, because it affects a minority of staff for a limited period and can be offset by structured mentoring. The saved commuting time, by contrast, is returned to every employee, every working day, for as long as the arrangement lasts."
Look at the machinery. The disadvantage is stated at full strength rather than weakened first, then weighed on two named grounds, who it affects and for how long, then the verdict is restated. That is what "outweigh" asks for, and it is the difference between band 6 and band 7 here. Band 7 asks for "a clear position throughout the response", with main ideas that are "extended and supported".
#Prompt B is not a weighing task at all
Prompt B asks two separate questions. The first is about causes, in the past. The second is a prediction, about the future. Nothing in it asks you to weigh good against bad, so an essay arguing that home working is a positive development has answered a question nobody asked.
The plan:
- Introduction. Paraphrase, then signal both answers in a single sentence.
- Body 1. Why it grew: two causes, one technological and one economic, each developed.
- Body 2. Whether it will last: a clear prediction, with the reason and the limit.
- Conclusion. Both answers, one sentence each.
The commonest way to lose marks here is unequal halves. Two hundred words on why it grew, forty words on whether it will continue. The band 6 descriptor names this precisely: the answer "Addresses all parts of the task although some parts may be more fully covered than others." Band 7 requires only that it "Addresses all parts of the task". So splitting your two body paragraphs evenly is not a style preference on a two part question. It is the descriptor.
Here is the opening of body 2, showing what a real prediction looks like:
"Whether the trend continues depends less on employee preference than on the cost of office space. Firms adopted home working because they had to, then discovered they could give up a floor of an expensive building. That saving is permanent and it appears on a balance sheet, which is why I expect hybrid arrangements to remain normal in office based industries for the foreseeable future. The limit is equally clear. Work that needs physical presence, such as manufacturing, healthcare and construction, cannot follow, so the shift will deepen in the sectors that have already made it rather than spread to new ones."
A prediction with a reason and a limit. Not "in the future, I think it will maybe continue because technology will improve."
#Worked example 2: public transport, two shapes
The same experiment on a different family. Two more real prompts from the bank:
Prompt C. "Some people argue that governments should invest more in public transport than in building and maintaining roads. To what extent do you agree or disagree?"
Prompt D. "More and more people are using private cars instead of public transport. What are the reasons for this trend? What measures could be taken to encourage the use of public transport?"
#Prompt C wants a position you never drop
"To what extent" is an invitation to say how far, not only whether. You have three honest options: agree, disagree, or agree with a stated limit. All three can reach band 9. What cannot reach band 7 is a position that drifts, where paragraph one supports buses, paragraph two supports roads, and the conclusion says both are important.
If you choose the limited position, say what the limit is in the introduction, and then honour it:
"I largely agree, with one exception. In dense cities, money spent on buses and trains moves far more people per pound than money spent on tarmac, so the priority is clear. In rural regions, where population density makes frequent services impossible to run, road spending is not a competing luxury but the only realistic form of transport investment."
That is a position, not a fence. The reader knows what you think and knows the boundary of what you think. Each body paragraph then defends one half of it, and the conclusion repeats the same shape. Coherence and Cohesion at band 7 asks you to "present a clear central topic within each paragraph", and a stated limit makes that almost automatic.
#Prompt D wants your opinion kept out of it
Prompt D is a problem and solution question, and it never asks what you think. Candidates lose marks here in a very specific way: they answer part one, then use part two to argue that cars are bad. That is a different essay.
The real skill in this type is that the two halves must line up. Every measure you propose should answer a reason you actually stated. Written out, the plan looks like this:
Reason 1: a car is faster door to door, because it has no walk at either end and no waiting. Matching measure: bus priority lanes and traffic signal priority, which attack journey time directly instead of attacking the fare.
Reason 2: the car is already bought, taxed and insured, so each extra trip feels free while each bus trip has a visible price. Matching measure: congestion charging, which puts a visible price back on the marginal car journey.
Now compare the version we see most often. The reasons given are "convenience" and "status". The measures given are "raise public awareness" and "build more trains". Nothing connects. Awareness does not answer status, and more trains does not answer a journey that was never going to use a train. The English can be flawless and Task Response still lands at 6, because the band 6 descriptor warns that "Main ideas are relevant but some may be inadequately developed or unclear". Solutions that do not answer the stated causes are the definition of underdeveloped.
Prompt D happens to be a General Training prompt, and GT Task 2 does sit closer to daily life than the Academic set. The marking is identical though: same four criteria, same descriptors, same 250 word minimum. If you are still choosing a paper, our guide to Academic versus General Training covers where the two tests genuinely differ, and Task 2 is not one of those places.
#The 30 second routine that prevents most of this
Before you plan anything, do this in order. It costs half a minute and it protects the criterion worth a quarter of your writing band.
- Find the instruction sentence. It is almost always the last sentence before "Write at least 250 words". Underline it.
- Count the questions in it. One question mark usually means one job. Two means two jobs and two body paragraphs of roughly equal size.
- Decide whether an opinion is wanted. Discussion, opinion and advantage questions want one. Problem and solution and cause and effect questions do not. Adding an opinion nobody asked for is untidy. Leaving one out when it was asked for is a Task Response penalty.
- Write your position, or your two answers, in six words each before you write any English. If you cannot state it in six words, you do not have it yet.
- Check the scope. If the prompt says "in your country", "for children" or "at university", every paragraph stays inside that boundary. Drifting outside it is the quiet way people lose half a band.
#Five mistakes we see most often, and the fix
- The memorised topic essay. A pre written free university essay bolted onto a prompt about university rankings. Fix: memorise argument moves, not paragraphs. The four moves earlier in this article are a starting set, and the habits that hold essays at band 6 covers the rest.
- Opinion smuggled into a problem and solution answer. Fix: after planning, reread the instruction and delete any sentence beginning "I believe" if the instruction never asked.
- The unequal two part answer. Fix: write the two body paragraph headings first and give each one a word budget, roughly 110 words each.
- Listing five reasons and developing none. Fix: two reasons, fully extended, beats five named. Band 7 wants ideas that are "extended and supported", which is the opposite of a list.
- Treating "to what extent" as "how strongly do you feel". It asks how far the statement holds, which invites a limit. A stated limit is usually the easiest route to a clear, sustainable position.
#How to practise this properly
Reading about question types will not change your band. Sorting real prompts will. Take ten prompts and do not write essays. For each one, write only the instruction sentence, the type, your position or your two answers in six words, and your two body paragraph headings. Ten plans in twenty minutes, then check them against the type list above. You are training the thirty seconds that decides the shape of the whole essay, and training it far faster than writing full essays would.
Then write two full essays a week, not ten, and get both marked properly. Our free IELTS writing checker scores all four official criteria and quotes the sentences holding the band down, which is what tells you whether the plan survived into the writing. If you want to know how far to trust an AI band before you plan around it, we publish our own scoring accuracy every week, including the weeks it looks bad.
One last thought on predicted topic lists. Even an accurate one would only hand you the half of the question you were already going to be fine with. The instruction sentence is the half that decides your band, and no list can predict it, because every topic in this article can arrive in any of six shapes.