Editorial standards

Everything here is original

No question, passage, audio script or prompt on this site is taken from a real CELPIP test. Official test material is confidential and copyright, and reproducing it would be both unlawful and useless, because a memorised item teaches nothing. Every item is written from scratch to match the published format: the part structure, the question counts, the timings and the question types.

What we match is the shape of the exam. Reading and Listening each carry 38 scored questions across their published parts. Writing has two tasks, Speaking has eight. Where the official format changes, the content changes with it.

How a test gets written

Each section has a written specification before any item exists. The specification fixes what a part contains, how many questions it carries, how long the candidate has, what the question types are, and what a correct distractor looks like. Items are then authored against that specification rather than freehand, so a test is a considered set rather than a pile of questions.

Content is written for a Canadian setting, because the exam is, and because a candidate should be practising the vocabulary and the situations they will actually meet: tenancy, municipal services, workplaces, community notices, transit, health care.

What is checked before a test ships

Human review catches the obvious. It reliably misses a specific class of error, so those are checked by script instead, and the scripts run over the whole corpus rather than one test at a time.

• Answer key order, not just its counts. A key can be perfectly balanced and still be guessable. One early reading test had a part whose key ran A A B B C C D D E, which is a free pass to anyone who notices by question four. The audit checks the sequence and the repeat runs, not only the totals.
• Answer distribution. A chi-square test per part and per test, so a key that quietly favours one letter is caught.
• Convergence across the set. Items written in parallel against the same specification drift towards the specification's own examples. One round of speaking tests used the same character name in five of six papers. Whole-corpus checks exist because nothing visible inside a single test is wrong in that case.
• Structural conformance. Part counts, question counts, timings and question types are checked mechanically against the specification rather than trusted.
• The build itself. Test banks are compiled into bundles the site serves. A stale bundle does not error, it silently serves a different paper under the label you asked for, so the build freshness of every bundle is verified before anything can be committed.

Alongside those, every published page has to satisfy a checked contract before it can be committed: a canonical URL, a unique title and description, one h1, one main landmark, valid structured data, and no mismatch between a visible FAQ and the structured data that describes it. A page cannot ship saying one thing to a reader and another to a search engine.

Where facts about the official exam come from

Anything this site states about the real CELPIP test, the format, the number of questions, the timings, the fees, how long a result stays valid, comes from Paragon's published material at celpip.ca, or from IRCC where the claim is about immigration. Score conversions to the Canadian Language Benchmarks follow the official comparison chart.

Facts that move are dated on the page that states them, because fees and requirements change and an undated number is a trap. Where a figure is a rule of thumb from our own data rather than a published one, it is labelled as such.

celpipmocks is an independent study site and is not affiliated with Paragon Testing Enterprises. We do not have access to the official marking engine, the official item bank, or any unpublished information about the exam, and nothing here should be read as though we do.

Model answers and sample responses

Model answers on the site are written to demonstrate a specific band and are accompanied by the reason they reach it, criterion by criterion. They are not templates. A memorised structure dropped onto a different prompt costs marks under Task Fulfillment, and the sample pages say so rather than pretending otherwise.

When something is wrong

Items are wrong sometimes. A distractor turns out to be defensible, a passage does not support the key, an audio clip and its transcript disagree, a conversion table goes out of date.

Report it to admin@celpipmocks.com, or through the contact form. Say which test, which part and which question number, and what you think the answer should be. What happens next:

• The item is checked against its source and its specification.
• If the key is wrong it is corrected, and the audits are re-run over the whole section rather than that one test, because a mistake of that kind is rarely alone.
• If the item is ambiguous rather than wrong, it is rewritten. An item with two defensible answers is a bad item even when the key is technically correct.
• If the item is right, you get an explanation of why, not a form reply.

Corrections change future scoring. They do not retroactively rewrite a report you have already received, because that report is a record of what you sat.

What we will not do

• Publish material copied or reconstructed from a live CELPIP paper, or accept it from anyone.
• Claim a band here is an official score, or imply an affiliation that does not exist.
• Inflate practice scores. A band that flatters you is worse than useless when the real test decides an application.
• Promise a score outcome. No preparation site can, and any that does is selling something.