Free CELPIP writing checker
This free CELPIP writing checker runs instant checks on a Task 1 email or a Task 2 survey answer: word count, paragraphs, greeting and closing, repeated words, very long sentences and phrases that sound memorized. These checks show what to fix, not a level. For an AI level estimate with feedback on all four things raters score, sign in to answer a writing task in practice.
Check your answer with the CELPIP writing checker
The heater in your apartment stopped working three days ago, and the nights are getting colder. In about 150–200 words, write an email to your building manager. In your email, you should:
- explain the problem
- say how it is affecting you
- ask for a repair date
What the instant checks look at
- Length against the 150–200 word range
- Paragraphs, a greeting and a closing
- Words used three times or more
- Very long sentences
- Stock phrases that sound memorized
Want a level?
Instant checks are not a level. For an AI level estimate with feedback on the four things raters score, answer a Task 1 prompt in practice. A free account includes 3 AI grades, then 1 a day.
Get my level estimatePractice estimate, not an official score.
Pick Task 1 or Task 2, then choose a prompt from our list or paste your own. Write or paste a draft, then select Check my writing. Every check runs in your browser, so no account is needed.
These checks look at shape: length, layout and a few weak habits. No check can judge whether ideas convince or grammar is right, so none of them shows a level. Fix what they find first, then ask for an AI estimate once your draft is as good as you can make it.
What each instant check looks for, and how to fix it
- Task 1 and 2
Word count
This check counts words against a range of 150–200. Under 150 usually means a point needs another detail or example; over 200, cut repeated ideas before you cut details.
- Task 1 and 2
Paragraphs
This check counts blocks of text with a blank line between them: at least four in an email, three in a survey answer. Put a blank line after a greeting, between points or reasons, before a sign-off.
- Task 1 only
Greeting and closing
This check wants a greeting on line one, such as “Dear Ms. Lam” or “Hi Joel”, plus a sign-off in the last two lines, such as “Kind regards” or “Take care”. Add whichever is missing, in a matching tone.
- Task 1 and 2
Your prompt’s points
After checking, tick off what was asked: every point of an email prompt, or a clear choice, two reasons and a line on the other option in a survey answer. If no sentence clearly does it, it is missing or too thin.
- Task 1 and 2
Repeated words
This check flags words of four letters or more used three times or more, leaving out common words such as “there” or “would”. Swap a repeat for a more exact word, or rewrite that sentence so it is not needed.
- Task 1 and 2
Very long sentences
This check flags any sentence over 35 words, quoting how it starts. Split it where a new idea begins, often at a linking word such as “and”, “because” or “which”.
- Task 1 and 2
Phrases that sound memorized
This check spots set lines that could start or end any answer, such as “I am writing this email to bring to your kind attention” or “last but not least”. Say it again in plain words about this situation.
“My heater stopped working on Monday. I called your office twice and left a message, but nobody has called back. Meanwhile, my apartment gets colder every night.”
“My heater stopped working on Monday, and I called your office twice, but nobody answered, so I left a message, and I am still waiting for someone to call me back while my apartment gets colder every single night.”
“My heater stopped working on Monday. Since I work from home, my cold apartment makes it hard to focus.”
“There is a problem with my heater. This problem started on Monday. This problem is serious because I work from home.”
Topic words can repeat for good reason: an email about a heater will say “heater” more than twice. A repeat check is most useful on general words such as “good”, “problem” or “things”, which usually have a more exact replacement.
Passing every check means a draft is well shaped, not that it has reached a level. Two drafts of identical length and layout can sit several levels apart, because ideas and language are what differ.
CELPIP writing evaluation criteria: four things raters score
Raters score every response, in both tasks, on four criteria. Here is each criterion in plain words, plus what earns credit.
- Content/Coherence: how many relevant ideas you develop and how clearly they connect. Raters reward specific details or examples, ideas in a logical order, one clear point per paragraph.
- Vocabulary: whether word choices fit the reader and situation, and how precisely you use them. Exact words (“a partial refund”, “out of order since Saturday”) beat general ones (“money back”, “not good”), while range should sound natural rather than forced.
- Readability: how easy a response is to follow. That means clear paragraphs, linking words showing how ideas connect, plus control of grammar, spelling and punctuation.
- Task Fulfillment: whether you did what a prompt asked. In Task 1, this means covering every point in a tone suited to its reader; in Task 2, choosing one option and supporting it. Staying inside 150–200 words is part of it.
All four criteria overlap. A missing point weakens Task Fulfillment and Content/Coherence at once, while a paragraph full of errors makes a good idea harder to follow. That is why quick gains usually come from content: one specific detail per point, in a fitting tone.
Get a CELPIP AI score estimate on a practice task
For a level, sign in to write a Task 1 practice email or a Task 2 practice answer under its own timer, then submit it for grading. A free account includes 3 AI grades, then 1 a day once those are used up. Every grade shows:
- an estimated level range, plus how confident our grader is
- a score and a comment for each criterion, with short quotes from your draft
- corrections of specific sentences, each with its reason
- three tips for reaching a higher level, plus a model answer to compare with
Why you see a range, not one number
A practice grade is not an official score, so your estimate shows where your writing most likely sits as a range, with a confidence label. That range is wider when our grader is less sure, for example when a draft is very short. Practice estimate, not an official score.
Our grader is AI only: no person reads your draft. CELPIP Trainer is an independent practice site, not affiliated with Paragon Testing Enterprises, which runs the official test.
How to use AI feedback well
Rewrite one paragraph at a time, starting from whichever comment would change most. Feedback helps when you act on it in a new draft; reading it once and moving on changes little.
- 1
Start with Task Fulfillment
If a point is missing or thin, fix that first. A new detail in a well-chosen place can lift two criteria at once.
- 2
Rewrite one paragraph, not everything
Pick the paragraph your feedback quotes most. Rewrite it using those corrections, then read old and new versions side by side.
- 3
Turn repeated errors into a short list
If a kind of error appears twice, such as a missing article or a wrong tense, write it down. Check for it in every draft you write next.
- 4
Use a model answer for its moves
Look at how it opens, how it supports each point and how it ends. Borrow its moves, not its sentences.
- 5
Grade fresh drafts, not a polished copy
Apply your list to a new prompt before your next grade. A grade on a new draft tells you more than a grade on one you have already corrected.
For models at three levels side by side, read our CELPIP writing samples. For a plan to follow while you rewrite, use our writing templates.
Before you submit: a two-minute check
- Task 1: every point listed is answered, each with a specific detail.
- Task 2: one option is chosen up front, then backed by two reasons with an example each.
- Word count: between 150 and 200.
- Task 1: greeting, tone and sign-off all suit the reader.
- No general word is repeated where a more exact one would do.
- Any sentence you could not read aloud in one breath is split in two.
- Verb tenses, articles and plurals are checked one last time.
If email writing is new to you, our Task 1 email guide works through one prompt with three model answers plus a plan for its 27 minutes. When both tasks feel comfortable, write them in our free mock test, after listening and reading. Spoken answers have their own page: our speaking checker.