Paragon publishes a surprising amount for free, and almost nobody links to it — including the prep sites charging you for worse versions of the same thing. The study packs are where the per-part detail actually lives. The Speaking pack carries the preparation and speaking time for each of the eight tasks, which the test-format page does not print. The Listening and Reading packs describe each part's audio and texts in the exam's own words.
Then there is the honest half. Free official material tells you the format. It cannot tell you whether you are getting better, because nothing in it marks you.
The failure
Two mistakes, opposite directions.
The first is paying for a repackaged copy. A "task breakdown" sold to you as premium content is often the study pack, retyped. The second is worse: treating the free official library as practice. You can read every PDF Paragon publishes, understand the format completely, and still have no idea whether your Speaking answer lands at 7 or at 9 — because you have never had one rated.
The index
| What it is | Genuinely best for | Limitation |
|---|---|---|
| Test format page | The canonical part names, question counts and section timings | Publishes no per-task Speaking timing and no task descriptions |
| Listening Pro Study Pack | Per-part speakers and clip lengths; the one-listen rule | Describes the parts, does not test you on them |
| Reading Pro Study Pack | What each part's texts are, including the "not stated" option | Pre-2023; its section timing is out of date |
| Speaking Pro Study Pack | Preparation and speaking seconds for all eight tasks | Not a rating — you still speak into nothing |
| Performance Standards PDF | The four rating dimensions for Speaking and Writing | Names the dimensions; publishes no weighting for them |
| Busting CELPIP Myths | The accent statement, in Paragon's own words | Short, and it is marketing-adjacent |
| Score Guidance PDF | Level descriptors, and "there are no points deducted for wrong answers" | No published raw-score cutoffs |
| Test results page | How many raters see your response; raw score out of 38 | Does not resolve how your two Writing tasks become one level |
The Speaking pack is the one to open first
The format page gives you 15 minutes and eight task names — "Giving Advice", "Talking about a Personal Experience", "Describing a Scene", "Making Predictions", "Comparing and Persuading", "Dealing with a Difficult Situation", "Expressing Opinions", "Describing an Unusual Situation". It stops there. The per-task clock lives in the study pack, which is why almost nobody publishes it.
| Task | Preparation | Speaking |
|---|---|---|
| 1 Giving Advice | 30 s | 90 s |
| 2 Personal Experience | 30 s | 60 s |
| 3 Describing a Scene | 30 s | 60 s |
| 4 Making Predictions | 30 s | 60 s |
| 5 Comparing and Persuading | 60 s prep with no speaking, then 60 s prep | 60 s |
| 6 Difficult Situation | 60 s | 60 s |
| 7 Expressing Opinions | 30 s | 90 s |
| 8 Unusual Situation | 30 s | 60 s |
Two tasks give you 90 seconds and the rest give you 60. Task 5 has a shape of its own. That is a real planning difference, and it is free.
The Listening pack, for what the audio is actually like
The format page says Listening is 46–55 minutes and 38 questions across six parts. The pack tells you what you will hear: Part 5 is a video with three speakers, Part 6 is one speaker for about three minutes, paper and pen are provided, and the rule that governs everything —
"All audio clips are played one time only."
The Reading pack, for structure only
It describes each part's texts: Part 2's main text is a diagram, and Part 3 offers a "not stated" option alongside the answer choices, which is the part where careful readers lose marks by inferring. Part 1 is 11 questions split 6 + 5 across its texts, Part 4 is 10 split 5 + 5.
The catch is in the URL — it was uploaded in 2022, and the section length it states is contradicted by the current site, which gives Reading 43–56 minutes. Use the Reading packs for structure, not for timing. Take every number that involves a clock from the format page.
The Performance Standards PDF, for what raters look at
Four dimensions for Speaking: Content/Coherence, Vocabulary, Listenability, Task Fulfillment. Four for Writing: Content/Coherence, Vocabulary, Readability, Task Fulfillment. Three of the four are shared — only the fourth changes with the medium.
Pair it with the myths page, which states plainly:
"Our raters specifically do not consider your accent when rating your responses."
But read the rubric before you relax. "Rhythm, pronunciation, and intonation" are Listenability sub-points, and the level descriptors talk about pronunciation errors impeding communication. Accent is not rated; being understood is.
The free online sample test, for the interface only
The one thing no PDF can give you: what the screen looks like. Open it to meet the layout, the timer and the answer controls before test day rather than on it.
Then read this limitation, because it is the one that can teach you a bad habit. On the sample test, "users have the ability to go back and review or change their answers." The real test does not allow that between parts — "You can't go back to a previous part." Practise the sample test's navigation and you are rehearsing a freedom you will not have.
The Guidebook for Test Takers, for the overview
The single document that covers the whole test at once — components, section timings, the score chart. Best for: one read-through when you are starting, so the individual packs make sense afterwards. Limitation: it is broad by design, so it will not answer a question about a specific part.
Guidebook for Test Takers, July 2026
What free official material cannot do
- Nothing in it marks you. Not one page, PDF or pack returns a level on something you wrote or said.
- No feedback. The Performance Standards name the dimensions. They do not tell you which one is costing you, and Paragon publishes no weighting for them.
- No answer to "is this a 7 or a 9?" You can read every descriptor and still not place your own response between two of them. That judgement is the thing you are missing, and it is the only thing that changes a score.
- No cutoffs. Listening and Reading come back as a raw score out of 38, and CELPIP labels its own raw-score-to-level table "Approximate". The confident cutoff tables elsewhere are reverse-engineered by third parties and disagree with each other.
- Some of it is stale. The Reading packs predate the current timings. Check anything time-related against the format page before you build a plan on it.
Read the free material once, properly. It is better than most of what is sold. Then stop reading and get something marked.
Try one
[Try a marked task →] — one full task, marked, no card required.