CELPIP Writing Task 1: what the rater is actually marking
Aman Batth · 12 min read · · Updated
Sample answers show you a finished email. They do not show you why it scored, which is the only part you can reuse. Task 1 is scored on four dimensions, and register leaks into three of them — which is why writing to a landlord the way you would write to a friend is a content failure, not a style preference.
Search for CELPIP Writing Task 1 and you will find pages of sample emails. Read ten of them and you will have ten finished products and no method, because the one thing a sample cannot show you is why it scored.
That matters more on Task 1 than on most writing tasks, for a reason that is specific to CELPIP. Paragon Testing Enterprises, which runs the test, says CELPIP writing is scored by “an AI-human hybrid system” in which all writing scores are “reviewed and verified by our rating team” (CELPIP FAQ). A human rater signs off on your email, and the thing they are weighing hardest is not your grammar.
This is written from the marking side — we build the engine that scores CELPIP writing practice on Hilingo. For the official criteria themselves the source is CELPIP's own performance standards, not a blog post.
What the task actually is
27 minutes. 150 to 200 words. One email. Writing Part 2, the survey response, gets 26 minutes after it, and the two tasks together produce one writing level. There is no separate Task 1 score on your report.
That last point is the one people miss. A weak email does not cost you "half of writing" — it is averaged into a single level with your survey response, so the two tasks are competing for the same number.
The prompt gives you three things, and each of them is load-bearing:
- A scenario — the situation you are writing about.
- A recipient — who the email goes to. Usually one clause, easy to skim past.
- Bullet points — what the email must include, listed explicitly.
Most candidates read the scenario carefully, read the bullets carefully, and skim the recipient. That is precisely backwards, and the next two sections explain why.
The four dimensions
CELPIP writing is assessed on content and coherence, vocabulary, readability and task fulfilment. Speaking uses the same four with listenability in place of readability. Every one of them is scored on the 12-level scale, and the levels map one-to-one to CLB — a CELPIP 9 is CLB 9. (The full mapping is in our free CELPIP score chart.)
In plain terms, here is what each one is asking:
| Dimension | The question behind it |
|---|---|
| Content and coherence | Does this email do its job, and does it hold together as one piece of writing rather than a list? |
| Vocabulary | Is the word choice accurate and suited to this reader — not merely rare? |
| Readability | Can a reader move through it easily: sentence control, punctuation, spelling, paragraphing, layout? |
| Task fulfilment | Did you do what the prompt told you to do, all of it? |
Notice what is not on that list. There is no criterion called "tone", "register" or "formality".
Why register is a content failure, not a style preference
Because register has no box of its own, candidates treat it as decoration — something you add if there is time. The opposite is true. A wrong register does not lose you a "style" mark; it damages three of the four dimensions at once.
- Content and coherence. An email's job is defined by its reader. A complaint to a building manager that reads like a message to a friend has not made the complaint — it has described a problem to someone with no reason to treat it as one. The content did not do its job.
- Vocabulary. Appropriateness is part of accuracy. "Hey, the heating's been rubbish for ages" is not a vocabulary gap; it is vocabulary aimed at the wrong person, which is a vocabulary error.
- Task fulfilment. The prompt named the recipient. Writing past them is not answering the task.
So an email to a landlord and an email to a friend are genuinely different tasks, not the same task in different clothes. This is one of the few places in language testing where getting the audience wrong is scored as getting the question wrong.
Our own CELPIP email prompts are built with this baked in: every Task 1 item carries a named recipient and a declared register — formal, semi-formal or informal — and the register has to match the recipient before the question is allowed into a paper. When the answer goes to the marker, the marker is told both.
What register actually changes
Register is not the greeting. Changing "Dear Mr Patel" to "Hi Sam" and leaving the rest identical is the most common half-attempt, and a rater sees it immediately.
| The prompt sends the email to | Register | What has to change through the whole email |
|---|---|---|
| A building manager, a company, a council department, an employer you have not met | Formal | Title and surname, no contractions, complete sentences, an explicit purpose sentence, a stated request, a formal sign-off |
| A colleague, a neighbour you know, a club organiser, a business you deal with regularly | Semi-formal | First name, some contractions, direct and warm, polite but not stiff |
| A friend, a family member, a classmate you are close to | Informal | First name, contractions throughout, conversational connectors, short sign-off, personal reference to your shared history |
The check: swap the greeting on your email for one from a different row. If the rest of the email still reads naturally under the new greeting, you never changed register — you changed one line.
Task fulfilment: the bullets are the mark scheme
The bullet points in the prompt are not suggestions about what to include. They are a checkable list, and each one is a discrete thing the rater can find or fail to find.
A missing bullet is the single most expensive, most avoidable loss in Task 1. It is not compensated by anything, because no amount of vocabulary answers a question you did not answer. An email that covers two of three bullets beautifully is an email that did two thirds of the task.
This is also the one error a candidate can eliminate completely, with thirty seconds of effort:
The check: before you write the final sentence, go back to the prompt and point at the sentence in your email that covers each bullet. If you cannot point at one, the bullet is not covered — and "I sort of implied it" does not count, because a rater marks what is there.
Content and coherence: the reader must know why you wrote within one sentence
A rater reads dozens of these. The emails that read as coherent tend to have the same shape, and it is not a template — it is what a purposeful email does naturally:
- Purpose first. One sentence that states why this email exists. "I am writing to report a recurring problem with the heating in flat 4B."
- The detail grouped, not listed. The bullets go into paragraphs in a logical order, connected to each other. A paragraph per idea, not a paragraph per bullet mechanically stamped out.
- A clear ask at the end. What do you want the reader to do? An email that never asks for anything is an email whose content has not landed.
Coherence is the ideas following from each other, not the presence of linking words. Starting every paragraph with "Furthermore" is signposting, and a rater reads mechanical signposting as exactly what it is.
Vocabulary: appropriate beats rare
The instinct under pressure is to reach for impressive words. On Task 1 this backfires twice over.
An unfamiliar word used slightly wrong damages vocabulary more than a plain word used correctly — that is true on any test. But CELPIP adds a second trap, because vocabulary is judged against the reader. Formal vocabulary in an informal email is also a mismatch. A message to a close friend that says "I would like to express my sincere appreciation" is not scoring higher for effort; it is scoring as language aimed at the wrong person.
Range within words you control, pitched at the reader the prompt gave you. That is the whole instruction.
Readability: the shape of the thing
Readability is the dimension candidates ignore because it feels like presentation. It is not. It is whether a reader gets through the email without friction — sentence control, punctuation, spelling, and the visible shape on the page.
Three paragraphs of readable length beat one block of 180 words, every time, and it costs nothing. A wall of text signals to a rater before they have read a single clause.
The word count is a constraint, not a target
150 to 200 words. Both ends matter, for different reasons.
Under 150 is the real risk. A short email has not produced enough language for a rater to see range in, which caps the level regardless of how clean the writing is. You cannot demonstrate what you never wrote.
Over 200 costs you somewhere else. The clock is the point: 27 minutes for the email, 26 for the survey response, and the two tasks share one level. Time you overspend on Task 1 comes directly out of Task 2 — and it is Task 2 that most often gets abandoned half-finished.
How to use a sample answer so it actually teaches you
Sample answers are not worthless. They are worthless read as prose, because you absorb the fluency and learn nothing transferable. Read them as evidence instead:
- Map every sentence to a bullet. Write the bullet number beside each sentence. Sentences that map to nothing are the email's connective tissue — notice how few there are.
- Highlight the register markers. Every contraction, every salutation choice, every softened request ("would it be possible to" versus "can you"). Now ask what each one would become if the recipient changed.
- Find the purpose sentence. It is almost always the first or second. Time how long it takes you to know why the email was written.
- Count the paragraphs, then the words. See what 170 words looks like in your own handwriting or typing, so you stop counting mid-exam.
Do that to three samples and you have a method. Read thirty and you have read thirty emails.
The 27 minutes
| Minutes | What you are doing |
|---|---|
| 0–3 | Read the scenario. Underline the recipient. Decide the register before anything else. Number the bullets. |
| 3–5 | One line per bullet, in the order you will write them. This is the plan, and it is enough of one. |
| 5–22 | Write. Purpose sentence first, then a paragraph per idea, then the ask. |
| 22–27 | Point at each bullet in your text. Check the greeting and sign-off match the register. Check the word count. Fix spelling. |
The last five minutes are not optional polish. Every check in them catches an error that costs a dimension.
Get it marked against the four dimensions
The reason self-review plateaus on Task 1 is the same reason it plateaus everywhere: your email makes sense to you. You know who you were writing to, so you cannot see that the email does not.
Hilingo gives you one full CELPIP mock free, scored, with the complete report and no card. Writing comes back as a level on CELPIP's 12-level scale with the four dimensions reported separately — content and coherence, vocabulary, readability and task fulfilment — so a low writing level tells you which one pulled it down rather than leaving you to guess. Every Task 1 prompt is built with a named recipient and a declared register, exactly as the real test frames it.
One thing to know about the real test: Paragon says every writing score is reviewed and verified by its rating team before it reaches you. Hilingo gives you the level instantly, dimension by dimension. Use it to find which dimension is weak, then fix that. That is exactly what it is built for.
- How CELPIP is scored — levels, CLB, and what a 9 actually needs
- CELPIP score chart — CELPIP level to CLB, IELTS and PTE Core
- PTE Core vs CELPIP — how much human review each test's writing gets, and what changes
- CELPIP vs IELTS vs PTE Core — which test is hardest in your weakest skill
Frequently asked questions
How long do I get for CELPIP Writing Task 1?
27 minutes for the email, and 26 minutes for the survey response that follows. The two tasks produce a single writing level, so time overspent on the email is taken directly from the task that has to earn the same number.
How many words should a CELPIP email be?
150 to 200. Going under is the bigger risk, because a short email does not give a rater enough language to see range in, which caps the level however accurate it is. Going over mainly costs you time you need for Task 2.
Does CELPIP Writing Task 1 have to be formal?
No — it has to match the recipient the prompt names. An email to a building manager is formal, one to a colleague or neighbour is usually semi-formal, one to a friend is informal. Using the wrong register is scored as a content problem rather than a style preference, because an email's job is defined by who is reading it.
What happens if I miss a bullet point in the CELPIP email?
It costs you on task fulfilment, and nothing else compensates for it. Vocabulary and grammar cannot buy back a requirement you did not address. Checking every bullet against a sentence in your own text is the cheapest mark in the whole section.
Is CELPIP writing marked by a computer or a person?
Both, and the split is the point. Paragon Testing Enterprises states that CELPIP writing is scored by an AI-human hybrid system and that all CELPIP writing scores are reviewed and verified by its rating team (Paragon's CELPIP FAQ), which is the main reason results take a few business days rather than one or two. PTE is a hybrid too, but a narrower one: Pearson publishes human expert review of the Content trait on certain tasks only, with traits such as Pronunciation and Oral Fluency machine-scored. The difference between the tests is how much human judgement is in the loop and on what, not people against machines.
Are CELPIP Writing Task 1 sample answers useful?
Only if you take them apart. Read as prose they teach you almost nothing, because you cannot see why they scored. Map each sentence to a bullet, highlight the register markers, and find the purpose sentence — three samples read that way are worth thirty read straight through.
What are CELPIP writing responses scored on?
Four dimensions: content and coherence, vocabulary, readability and task fulfilment, each on the 12-level scale that maps directly to CLB. There is no separate criterion for tone, which is exactly why a register mistake is expensive — it damages three of the four at once.