CELPIP Writing Task 2 Template: How to Respond to a Survey

You get 26 minutes and 150 to 200 words. Task 1 gets 27 minutes, so the two tasks are not the same length, and the whole Writing component runs 53 minutes.

Those numbers come from Prometric directly. Its CELPIP-General Overview Study Pack, states "26 minutes to complete" and gives the length as 150 to 200 words for Task 2, and its Writing Pro: Target 9+ Study Pack puts it plainly: you have 27 minutes for Task 1 and 26 minutes for Task 2.

On word count, people want a hard number for how far they can stray. There isn't one. Prometric's published guidance is that you have "a little bit of room to be under or over," but your score "may suffer if you are significantly over or under." Any blog quoting a precise buffer like 135 to 220 words invented it.

Here is the more useful anchor: aim for about 180 words. A strong answer does not need to reach 200, and the extra 20 words at the end are usually where people repeat themselves and make mistakes.

The screen gives you basic editing tools, an automatic word counter, and a spell check. Prometric does warn that "it is sometimes difficult to choose the right word when the spelling checker offers you a choice," which is a polite way of saying the spell check will flag a word and then hand you five options with no help picking one. It catches spelling only. It will not catch affect where you meant effect, and it will not catch a verb that disagrees with its subject.

Four parts: your choice, reason A with an example, reason B with an example, and a close that acknowledges the other side. Four paragraphs, one blank line between each.

That layout is doing work for you before the rater reads a word. Paragraphing sits inside Readability, one of the four scored dimensions, and a rater who has to hunt through a wall of text for your structure is a rater you have already made work harder than necessary.

Two of those four parts are not just my preference. CELPIP's own Task 2 guidance says to open with your opinion, close with a concluding statement, and compare your choice with the other option. Parts 1, 4 and the nod are the test owner's advice, not a template someone invented.

Where reasons come from

You need two reasons that genuinely differ. "It helps families" and "it helps parents with young children" is one reason written twice, and it costs you under Content/Coherence.

Pick any two rows from this grid, as long as they are different rows.

Where examples come from

This is the step people freeze on. You have never lived in a building with a delivery policy, so what do you write?

You are not being fact-checked. Prometric's guidance is to "mention real places, people, and situations to make details more concrete and vivid," and a small plausible detail does that job perfectly well. What is being scored is whether you can turn an abstract claim into something specific.

A detail counts if it has at least one of these: a time, a number, a place, or a specific person.

Try this on a real Task 2 question, free

Sign up free and answer the first Writing Task 2 survey question on the real clock. You get up to three AI-scored tries, with feedback on what to fix before test day.

Answer a free Writing Task 2 question

Two-option prompts ask you to pick between A and B. Agree/disagree prompts give you one proposal and ask whether you support it. The template does not change between them.

Most preparation only covers the first shape, which is fine right up until you meet the second one on test day with 26 minutes on the clock.

Only two things differ.

Everything in between is identical. Learn one map and read the prompt carefully enough to know which shape you are in.

A note on where this comes from: Prometric describes Task 2 as choosing between two options. The agree/disagree framing is my own observation from the prompts my students actually report seeing, not something Prometric publishes. Prepare for both and you cannot be surprised.

Prometric rates Writing on four dimensions: Content/Coherence, Vocabulary, Readability, and Task Fulfillment. There is no points formula and no fixed number of marks per dimension. Prometric publishes the four and what sits inside each one, and does not publish weights.

That table is our own summary of CELPIP's published Writing performance standards. The official version is on celpip.ca.

Two things in it surprise people.

Grammar is not its own score. It sits inside Readability, alongside spelling, punctuation and paragraphing. A grammatically perfect answer with no paragraph breaks and no transitions can still lose Readability marks.

Word count is scored under Task Fulfillment. Not as a separate penalty, but as one of four factors alongside relevance, completeness and tone. Being significantly short is a Task Fulfillment problem, the same category as ignoring half the prompt.

For how these four dimensions translate into a CELPIP Level and a CLB (Canadian Language Benchmark) number, see how CELPIP Writing scores map to CLB. CELPIP Levels and CLB levels run one to one from 2 through 12.

Neither. Prometric's Task 2 strategy sheet gives two instructions that most prep material ignores: "Don't write an academic essay," and, on format, "No need to use email format here."

Both appear in Prometric's own published study materials, the first in the Writing Pro: Target 9+ Study Pack and the second in the 2026 CELPIP-General Overview Study Pack. Read them together and the target register is clear. You are a person answering a survey about your own street, your own building, or your own workplace. Not a consultant filing a report.

This matters more than any template line, because the essay register is the single most common way a strong English speaker caps their own score. An answer can be grammatically flawless, correctly structured and exactly 180 words, and still stall, because it reads like a memo instead of a person.

"Temperature degradation" just means the food gets cold. The longer phrase is not more advanced. It is only heavier, and reaching for heavy words at speed is how small errors creep in. Vocabulary is scored on precision and fit, not difficulty.

The padding diagnostic

Run this over your own draft. Every pattern below is a symptom of writing to impress rather than to communicate.

The test: read your sentence out loud. Would you say it to a friend describing the same situation? If not, rewrite it in the words you would actually use.

On greetings and sign-offs

Prometric says email format is not needed. That is not the same as forbidden, and a greeting will not cost you marks on its own.

My recommendation is still to skip it. Open with your opinion instead. You have 180 words, a greeting and a sign-off cost you roughly ten of them, and they buy you nothing on any of the four dimensions. Spend those ten words on a concrete detail instead.

Six answers below, three of each prompt shape. Each one states its word count and the two angles it uses, so you can see the machinery rather than just the output.

Read them for structure, not for phrases to memorize. A rater who has seen the same sentence four hundred times is not impressed by the four hundred and first, and Prometric's policy goes further than that: memorised or copied content can invalidate a result. Here is where the line sits between a structure and a script.

At every level the writer picks an option. What moves the score is how concrete the reasons are, how fairly the other option is treated, and how cleanly the sentences are built. Below is one survey question answered three times, as our examples of a Level 7, a Level 9 and a Level 10 response, each checked against Prometric's four rating categories.

These are our own practice answers, written to match Prometric's published Writing level descriptors (May 2026) and compared with the rated Task 2 responses (Levels 8, 9 and 10) in its Writing Pro: Target 9+ Study Pack. They are not official CELPIP responses, and no CELPIP rater scored them. CELPIP levels match CLB levels one to one.

My recommendation is 3 minutes planning, 18 minutes writing, and 5 minutes checking. Prometric's own recommendation is slightly different, and both work, so here is each one plainly.

Prometric suggests spending "around 4-5 minutes planning each response" and "around 3 minutes checking your response after you write." That is a real alternative from the test owner, and if you plan slowly and edit fast, use it.

I push the planning shorter and the checking longer for one reason: almost nobody loses marks for a thin plan, and almost everybody loses marks in the last paragraph they wrote in a hurry. Neither split is official. Prometric publishes no required breakdown of the 26 minutes, so this is a coaching choice and you should take whichever one matches how you work.

The 30-second rule on picking a side is the part I would not bend. There is no correct answer, and nobody is scoring your opinion. Ask only this: which side can I give two concrete examples for, right now? Choose that one, even if you personally prefer the other, and do not reopen the question at minute twelve.

The three that cost the most marks are treating it like an email, arguing both sides equally, and writing in an essay register. Here is the full list with the fix.

The closing nod, in three moves

The nod is the part people cut when they run short, and it is the wrong thing to cut. Prometric lists "compare your choice to the other choice" under convincing support, which means acknowledging the alternative earns credit under Content/Coherence and Task Fulfillment both.

Two or three sentences, never more. This is a closing note, not a second essay.

Compared with the previous version of this table, the padded openers are gone. "After careful consideration, I strongly support," "the most compelling reason," "cannot be overstated" and "I strongly urge" all fail the padding diagnostic above, so they are no longer here.

Vocabulary by topic, for when your mind goes blank at minute four. Use the plainest word that is precise, not the biggest one on the row.

The verb columns changed in this revision. "Optimize," "foster," "facilitate," "cultivate" and "revitalize" are gone, because they are the corporate-verb row of the padding diagnostic. A rater reading "optimizes municipal spending" sees someone reaching. A rater reading "saves the city money" sees someone who knows what they mean.