CELPIP Writing Task 2: the survey wants a choice, not a balanced essay
Aman Batth · 14 min read · · Updated
Task 2 looks like an opinion essay, so candidates write one — they weigh both options fairly, conclude that each has merit, and lose marks for it. The prompt is a survey with two options and it is asking you to pick one. Balance is not sophistication here; it is a failure to answer.
The most expensive mistake on CELPIP Writing Task 2 is made by people who are good at writing essays.
They read the prompt, see two options, recognise the shape from every academic writing class they have ever sat in, and produce a careful, even-handed piece: here is the case for the first option, here is the case for the second, both have merit, it depends on the circumstances. It is fluent. It is organised. It has not done the task.
Task 2 is officially called Responding to Survey Questions. A survey asks you which one you want. An answer that will not say is not a nuanced answer — it is a blank in the response field, written at length.
This is written from the marking side; we build the CELPIP writing scorer on Hilingo. For the official criteria themselves, read Paragon's performance standards rather than any blog, this one included.
What Task 2 actually puts in front of you
26 minutes. 150 to 200 words. The prompt gives you a short scenario — a city council consulting residents, an employer asking staff, a community organisation deciding how to spend a budget — followed by one statement and two options, labelled A and B. You choose one and explain why.
Two structural facts follow from that, and both are worth more than any vocabulary list.
You are a respondent, not an author. The scenario names a body that is asking, and you are one of the people it asked. That is a different job from arguing a thesis, and it is why "the question of urban green space is a complex one" is a sentence with nowhere to go. Nobody surveyed you about complexity.
It shares a writing level with the email. Task 1 and Task 2 do not receive separate scores — they are combined into one writing level on the 12-level scale, and CELPIP levels 3 to 12 map one-to-one to CLB 3 to 12, so a CELPIP 9 is CLB 9. (The full mapping is in our free CELPIP score chart.) The email task has its own trap, which is a different argument entirely: see CELPIP Writing Task 1.
The IELTS habit that costs you here
If you have prepared for IELTS, you have been trained on prompts where covering both sides is required. The discussion prompts used throughout IELTS preparation ask you to address two views, and an essay that ignores one of them is an incomplete answer. IELTS’ own published sample tasks include "Do the disadvantages of international tourism outweigh the advantages?", which demands that you actually weigh them, which many candidates fail to do — that failure is one of the IELTS mistakes you cannot see from inside your own answer.
So a well-drilled IELTS candidate arrives at CELPIP with a reflex that is correct on the other test and wrong on this one. The reflex says: cover both, then lean. The CELPIP prompt does not contain a proposition to evaluate. It contains a menu.
Here is the difference stated plainly, because it is the whole post:
| The prompt hands you | A complete answer | Covering both sides is | |
|---|---|---|---|
| IELTS Writing Task 2 | A proposition, sometimes with an explicit instruction to discuss two views | An essay that does what the instruction asked and holds a position through it | Often required by the wording |
| CELPIP Writing Task 2 | A survey statement and two options, A and B | A response that names one option and supports it | Not what was asked, and it spends the words you need |
Neither test rewards a fence-sitter. But on IELTS the fence has a gate in it, and on CELPIP there is no fence — there are two doors and you were asked which one you would use.
Why hedging is a task-fulfilment problem, not a style preference
CELPIP writing is rated on four dimensions: content and coherence, vocabulary, readability and task fulfilment (CELPIP performance standards). There is no dimension called "balance" and no dimension called "sophistication", which is why the even-handed response does not get credited somewhere for its fairness. It gets debited in two places.
Task fulfilment. The task said choose. You did not choose. Everything else in the response is language attached to an unanswered question, and no amount of accuracy buys back a requirement you did not meet.
Content and coherence. This is the one people do not see coming. A response that argues both ways has no line running through it, so its paragraphs are not connected by anything except the topic. Coherence is ideas following from each other; two cases for two options do not follow from each other, they sit beside each other.
Then there is the arithmetic, which is brutal and entirely avoidable. You have 150 to 200 words. Call it 170.
| How you spend 170 words | Words left for the option you chose | What the rater can see |
|---|---|---|
| Commit to one option, two reasons, specifics in both | ~140, once the opening and close are paid for | Two developed reasons with concrete support |
| Cover both options fairly, then lean at the end | Roughly half that, split across two thin reasons | Two half-reasons and a late preference |
| Cover both options fairly, never lean | — | No choice at all |
Balance does not merely fail the instruction. It halves your evidence on the option you actually preferred, which is the only evidence the response was supposed to produce.
What "support it concretely" means, and the test for it
Committing to an option is the easy half. The half that decides the level is what you put underneath it, and the failure here is subtler than hedging because it feels like writing.
A generic reason is a reason that would support either option equally well. It is not support; it is a sentence in the shape of support.
Run this check on every reason you write: swap the option and reread the sentence. If it still stands, delete it.
| Reason as written | Would it also support the other option? | Verdict |
|---|---|---|
| "This option would benefit the whole community." | Yes, exactly as well | Not a reason |
| "It would improve residents' quality of life." | Yes | Not a reason |
| "In today's fast-paced world, priorities are changing." | Yes, and for any prompt on any topic | Not even about this prompt |
| "The nearest supermarket is a twenty-minute bus ride from my street." | No | A reason |
| "There is already a park three blocks north with lighting and a path." | No | A reason |
The specifics in the bottom two rows do something the top three cannot: they make the choice follow from a situation rather than from a preference. That is what a survey response is for.
One thing candidates ask and almost nobody answers: the details do not have to be true. None of the four rated dimensions is about factual accuracy: content and coherence, vocabulary, readability and task fulfilment all judge the English you produce about a situation, not the situation. Invent the street, invent the bus. Then be consistent about it for all 170 words, because an invented detail that contradicts itself is a coherence problem even when the invention was fine.
A structure, and why it is not a template
Four moves. They are not sentences to memorise; they are jobs to be done, and the words change with every prompt.
- Name the option in the first sentence. Not hint at it. Name it — "Option B" plus what Option B is, in your own words, so the choice survives even if the labels are stripped.
- One reason per paragraph, each with a specific. Two developed reasons beat three thin ones at this word count. The specific is the part that does the work; the reason without it is a category.
- Handle the other option in a clause, if at all. You may acknowledge it — a brief concession can make your choice look considered rather than uninformed. It belongs inside a sentence that is still arguing your case, not in a paragraph of its own. The moment it gets its own paragraph, you are writing the balanced essay again.
- Close by restating the choice, not by summarising the topic. "For this lot, the store solves a problem the area actually has" is a close. "In conclusion, both options have their advantages" undoes the entire response in seven words.
Why a memorised template is worse than nothing here
A template is a set of sentences chosen before you saw the prompt. That is its definition, and it is also the whole problem, in a way that does not depend on anyone catching you.
The specifics are the only thing turning your reason into a reason, and specifics cannot be pre-written — they come from the option in front of you. So every templated sentence is a sentence that cannot contain support, spent out of a 200-word ceiling. A response that is forty percent scaffolding has forty percent less room for the thing being marked. That cost is arithmetic, not detection.
Detection is a separate matter, and worth being precise about. Paragon states that CELPIP writing is scored by an AI-human hybrid system combining artificial intelligence with its human rating panel, and that all CELPIP writing scores are reviewed and verified by that rating team (Paragon's CELPIP FAQ). We cannot see inside that system and will not tell you what it flags — nobody outside Paragon knows. What we can say is what follows from any marking process with a human in it: the same prompts recur, raters read many responses to each of them, and an opening sentence that arrives identically from several candidates is conspicuous to a reader in a way it is not to a candidate. You only ever see your own copy of it.
The same logic applies on other tests, incidentally, and it is why PTE templates do not work either.
A worked response, and what each part is doing
Here is a prompt of the kind Task 2 uses:
Your city council is asking residents what should be done with a vacant lot on Elm Street. Choose the option you prefer and explain your choice.
Option A: Turn the lot into a public park with benches and a playground. Option B: Lease the lot to a developer to build a small grocery store.
And a response, 170 words:
I am writing to support Option B, leasing the lot to a developer for a grocery store.
The main reason is distance. The nearest supermarket is a twenty-minute bus ride from Elm Street, so anyone without a car currently plans a whole afternoon around buying milk. A store on the lot would put fresh food within walking distance of the two apartment blocks at the north end, and the older residents there are the people who would notice the difference first.
The second reason is that this neighbourhood is not short of open space. Riverside Park is three blocks away and the school field is open to the public at weekends, so a second park would duplicate something we already have. A grocery store would duplicate nothing.
I understand the appeal of a playground, and if the council wants one I would rather it went into Riverside Park, where there is already lighting and a path. But on this particular lot, the store solves a problem the street actually has.
Now the annotation, which is the part worth copying:
| Where | What it is doing |
|---|---|
| Sentence 1 | Names the option and says what it is. The choice is unmissable in nine seconds. |
| "The main reason is distance." | Labels the reason before developing it. Cheap coherence, no connector needed. |
| "twenty-minute bus ride", "buying milk" | The specific. Neither of these sentences would survive a swap to Option A. |
| "the two apartment blocks at the north end" | Invented and consistent. It names who benefits, which turns a reason into a consequence. |
| "not short of open space" | The second reason attacks the other option's premise while still arguing for B. That is how you use the alternative without giving it a paragraph. |
| "I understand the appeal of a playground…" | The concession, in one sentence, ending with a redirect that is still arguing for B. |
| Final sentence | Restates the choice and the reason together. No "in conclusion", no summary of the topic. |
Notice what is absent. There is no introduction explaining what surveys are. There is no sentence beginning "There are many advantages and disadvantages to consider." Every sentence is either the choice, a reason, a specific, or the close — and that is what fits in 170 words.
The 26 minutes
| Minutes | What you are doing |
|---|---|
| 0–2 | Read the statement and both options. Choose. The one you can be concrete about beats the one you privately agree with. |
| 2–4 | Two reasons, one specific each, jotted as fragments. That is the plan and it is enough of one. |
| 4–21 | Write. Option named in sentence one, a paragraph per reason, the concession and close together at the end. |
| 21–26 | Swap-test each reason. Check the first sentence still names the option. Check the last sentence has not gone vague. Word count, then spelling. |
The choice in the first two minutes is the decision with the highest leverage in the whole task, and candidates agonise over it for the wrong reason. You are not being marked on which option you picked. Pick the one you can produce a street, a bus route and a consequence for. Conviction is not a criterion; support is.
Get the two tasks marked separately, then combined
The reason self-review fails on Task 2 is that your own response reads as decisive to you. You knew which option you had chosen before you started typing, so you cannot see that the text never said so plainly, or that both of your reasons would have worked for the other option.
Hilingo gives you one full scored CELPIP mock free, with no card. Writing comes back as a level with the four dimensions reported separately — content and coherence, vocabulary, readability and task fulfilment — so a hedged survey response shows up as task fulfilment sitting below your vocabulary, rather than as one number you have to interpret. There are 29 full CELPIP-General mocks and 28 CELPIP-General LS mocks in the library, and unlimited attempts on a paid plan. The AI teacher works from your own marked response, in English or any of 50 languages, so "why did task fulfilment come back low" is a question you can actually ask.
One caveat we would rather state than bury: Paragon says every real CELPIP writing score is reviewed and verified by its human rating panel. Ours runs start to finish without a human in it. Use our level to find which dimension is leaking, not as a prediction of the number you will be sent.
- CELPIP Writing Task 1 — the email, and why the recipient decides three of the four dimensions
- How CELPIP is scored — levels, CLB, and what a 9 actually needs
- CELPIP score chart — CELPIP level to CLB, IELTS and PTE Core
- Why PTE templates do not work — the same argument, run against a machine-scored test
- The IELTS mistakes you cannot see — the same blind spot, run against the IELTS criteria
Frequently asked questions
How long is CELPIP Writing Task 2 and how many words?
26 minutes, and 150 to 200 words. It follows the email task, which gets 27 minutes, and the two produce a single writing level between them — so time you overspend on the email comes out of the task that has to earn the same number.
Do I have to choose one option in CELPIP Writing Task 2?
Yes, and saying so explicitly is part of the task. The prompt gives you a statement and two options and asks which you prefer. A response that treats both fairly and never commits has not answered a survey question, which costs you on task fulfilment and leaves your reasons half-developed because the word count is split.
Can I mention the other option in CELPIP Task 2?
Yes, briefly, as long as it is still arguing your case. A single sentence acknowledging the appeal of the option you rejected can make your choice read as considered. Give it its own paragraph and you have written a balanced essay, which is the one thing this task does not want.
Is CELPIP Writing Task 2 the same as an IELTS essay?
No, and the resemblance is what makes it dangerous. IELTS Task 2 gives you a proposition to evaluate, and its discussion prompts ask you to address both views, so covering both sides is required there. CELPIP hands you a menu instead of a proposition — there is nothing to weigh, only something to pick — and a candidate running the IELTS reflex loses marks for doing the right thing on the wrong test.
Should I use a template for CELPIP Writing Task 2?
No, for a reason that holds even if nobody ever notices. A template is words chosen before you saw the prompt, so it cannot contain the specifics that make your reason a reason — and it spends part of a 200-word ceiling on scaffolding. Learn the four jobs instead: name the option, one reason per paragraph with a specific, a one-clause concession, a close that restates the choice.
Do the examples in my CELPIP survey response have to be true?
No. CELPIP writing is rated on content and coherence, vocabulary, readability and task fulfilment — none of which is about factual accuracy. Invent the details you need, then keep them consistent across the whole response, because an invented specific that contradicts itself later is a coherence problem even though the invention was fine.