CELPIP Speaking Tips: Strategies for All 8 Tasks (CLB 9+)

Speaking is the section most test-takers walk into already nervous. Headphones on, mic open, timer counting down. No examiner across the table, no nod when you say something clever. Just you and the recording.

Here's what should settle the nerves: the 8 tasks repeat. Same formats every test, similar topic pools, the same four scoring dimensions year after year. Practise each task type a few times and the surprises mostly disappear.

This guide walks through all 8 tasks with the time budget, the one move that lifts the score, and a link out to the deeper article for each. Plus the rubric-to-move mapping nobody else bothers with, and the specific moves that take a CLB 7 speaker to CLB 9 (or 10).

Two tasks give you 90 seconds of speaking time: Task 1 and Task 7. The rest give you 60. Tasks 5 and 6 trade extra speaking time for extra prep time, because the prompts hand you more information to process. Knowing the CELPIP speaking time per task lets you plan how much to say. Finishing at 45 seconds on a 60-second task leaves points on the table. Fill the time.

Content & Coherence

Does your answer match what the prompt asked, in a logical order? This is where most candidates lose points without realising. You can speak beautifully about the wrong topic and still drop a full band.

Spend the first 5 seconds of prep restating the question to yourself. Not the topic. The exact ask. If the prompt says give advice to a friend who can't decide between two job offers, your answer is advice. Not your own job-offer story.

Vocabulary

Vocabulary scores reward variety, not difficulty. You don't need exam-prep words. You need the right word used naturally.

Try this. Swap one general word per response. "A good restaurant" becomes "a packed neighbourhood place with terrible parking and great food." Specific beats fancy. That's the move you'll see Reddit threads credit for the jump from CLB 8 to 10: speak vividly, not academically.

Listenability

This one is about effort. Can the rater understand you without leaning in? Pronunciation and pacing live here. Accent doesn't. Mumbling does.

Drop the end-of-clause filler ("like", "you know", "sort of"). Replace each one with a half-second pause. The pause sounds confident. The filler sounds hesitant.

Task Fulfillment

Did you actually do what the task asked? Task 5 says compare and persuade. Did you compare AND persuade, or only describe? Missing one element drops the score noticeably.

In your 30-second prep, write the verb from the prompt at the top of your scratch paper. Advise. Predict. Compare. If your answer doesn't contain that verb's action, you missed the task.

The structure that fits 90 seconds: a 15-second setup acknowledging the dilemma, two recommendations with one reason each, then a short close that puts the choice back in their hands.

What raters reward

• Modal-verb stacking: "you might want to consider...", "I'd suggest...", "have you thought about...". Softening grammar grades higher than imperative "you should do X."
• A specific reason after every recommendation. "Take the job because it offers growth" beats "take the job, it's better."
• Treating the listener as a person, not a topic. "You know your family best" lands better than "family considerations matter."
• One concrete example: "my cousin took the same kind of offer two years ago, and..."

The deep dive lives in the dedicated Task 1 guide: annotated CLB 9 sample answer, modal-verb register chart, and the lecturing-tone trap most candidates fall into.

Sixty seconds is one short story, not three. Pick one scene with concrete details and stay there.

What raters reward

• Vivid, specific detail. Skip "a nice trip." Try "three days in Banff in February, snow up to my knees, a hot chocolate that cost twelve dollars." That's the one move test-takers on Reddit credit for the CLB 8 to 10 jump.
• Consistent past tense. Mixing tenses sounds confused even when the grammar is technically fine.
• Time-marker connectors. "Last summer", "by the time we got there", "eventually", "looking back." These anchor the story arc.
• Emotional vocabulary. Proud, relieved, embarrassed, surprised. One emotion word per response pulls the vocabulary score up.

You don't have to tell a true story. Raters aren't checking your life. They're checking your English.

Don't jump around the image. Pick a path and walk through it.

The quadrant method

In your 30 seconds of prep, split the image mentally into four quadrants: top-left, top-right, bottom-left, bottom-right. Note the main subject of each quadrant. When you speak, work through them in order. The structure handles itself.

What raters reward

• Present-continuous as the spine: "a woman is reading", "two children are running".
• Spatial prepositions: in the foreground, on the left, between the bench and the path, behind the food truck.
• One short observation about the mood: "the atmosphere looks relaxed", "everyone seems busy".
• Don't narrate intentions you can't see. "A man is sitting on a bench" is description. "A man decided to sit because he was tired" is invention.

This picture and the one in Task 4 are usually the same image with different questions. Set up your description in Task 3 so you can predict from it in Task 4.

Your predictions must connect to what's visible. If someone's holding an umbrella under cloudy skies, predicting rain works. Predicting a marriage proposal does not.

What raters reward

• Two or three distinct predictions with reasoning. One prediction can't fill 60 seconds without padding.
• Future and modal verbs in rotation: will, going to, might, could, probably. Mixing certainty levels grades higher than "will" repeated five times.
• Connecting back to the picture: "because the children look tired, they'll probably leave soon". The "because" earns the band.
• One slightly longer-term prediction: "by the end of the afternoon..." Stretches the time horizon and shows range.

The trap here is the fence. Don't sit on it. "Both options are good" doesn't show persuasion skill.

The structure

1. Acknowledge both options exist (10 seconds)
2. State your choice clearly (10 seconds)
3. Advantage 1 with reasoning (15 seconds)
4. Advantage 2 with reasoning (15 seconds)
5. Address a concern about your choice and counter it (10 seconds)

What raters reward

• Comparative language stacked through the response: "better than", "more reliable than", "less expensive than".
• Direct address to the listener: "you'll find that...", "this gives you...".
• An acknowledged downside, then a counter. "You might worry about the higher cost, but over a year it actually saves you..." Acknowledging the weakness strengthens the persuasion.

Use the full 60 seconds of prep. Read both options. Pick one. Write three bullets. Start speaking with a plan. The candidates who freeze on Task 5 are the ones who started reading and started speaking at the same time.

This task tests how you communicate under friction. Aggressive tone scores poorly even when the grammar is perfect.

Acknowledge, justify, resolve

Open by showing you understand the other side, even briefly. Justify your position with one specific reason. Close by proposing a concrete next step. That three-move arc handles almost every Task 6 prompt the test will give you.

What raters reward

• Diplomatic phrasing: "I appreciate you taking the time", "I'm hoping we can find a way to...", "I understand this wasn't intentional".
• A specific ask. "I'd like a refund" or "could we reschedule to Thursday?" beats vague complaint.
• Calm tone all the way through. If you sound annoyed at second 5, the rest of the recording feels off even if the content is fine.
• Empathy for the other side, then your position. Not the other way around.

Most candidates plan for 60 seconds and end up repeating themselves. Task 7 needs two real reasons with examples, not one reason stretched thin.

The 90-second structure

1. Clear stance up front (10-15 seconds): "I firmly believe..." Pick a side.
2. Reason one with a concrete example (25-30 seconds)
3. Reason two with a different example (25-30 seconds)
4. Acknowledge the other view briefly (15 seconds)
5. Restate the stance and close (10 seconds)

What raters reward

• Hedging vs commitment grammar used deliberately. "I tend to think" softens. "I'm convinced" commits. Mix both.
• Concrete examples over abstract argument. "In my neighbourhood, three cafes closed when remote work spiked" beats "remote work hurts local business."
• One "some people might argue... but I think..." structure to show you can see both sides.

Think of it as a phone call. The other person can't see what you're seeing. Your job is to put the picture in their head.

What raters reward

• Comparison phrasing: "it's shaped like a...", "it reminds me of a...", "it's about the size of a coffee mug but...". Comparisons rescue you when you can't remember the exact word for something.
• Two-axis description: physical attributes (size, shape, colour, material) and functional attributes (what it does, how you'd use it).
• Spotting the unusual element early. It's often not in the centre of the image. Scan the edges during prep.
• A short reaction at the end: "the strangest thing about this scene is...". Closes the response naturally.

CLB 7 to CLB 9: structure and specificity

A CLB 7 speaker can communicate. The grammar mostly works and the ideas come through. What's missing is shape. Listeners can follow the response, but the response doesn't lead them anywhere.

Three moves that take you from 7 to 9:

• Open with structure. Use the 30 seconds of prep to plan the shape (intro, two points, close) before you plan what to say. Speakers who freeze in prep usually started writing sentences instead of an outline.
• Swap one general word per response with a specific one. "A bad meeting" becomes "a 90-minute status meeting where four people read slides aloud." That single swap moves the vocabulary band.
• Use the verb from the prompt. If the task is "advise," your answer needs the word "advise" or a clear synonym in the first few seconds. Task fulfillment is the dimension most candidates ignore.

CLB 9 to CLB 10: range and naturalness

A CLB 9 speaker hits the rubric on every dimension. What separates 9 from 10 is range. The 9 speaker is consistent. The 10 speaker can flex inside a single response: a casual aside one moment, careful description the next.

The moves at this level:

• Vary sentence length deliberately. Long sentence. Then a short one. Then a medium one with a dependent clause. The rhythm sounds human. Five sentences all twelve words long sounds rehearsed.
• Mix register inside the response. Drop into casual phrasing for emphasis ("honestly, that's the part that surprised me"), then return to careful description. Range across the same 60 seconds reads as control, not inconsistency.
• Let the close land. Most candidates fade at second 55 because they planned 60 seconds of content and start running out. Plan 55 seconds of content and a deliberate closing line: "and that's why I'd take the second offer." The close is the last thing the rater hears.

On Celpify, the practice attempts that score CLB 10 in our AI rater almost all share one trait: the speaker slows down in the final ten seconds and the close is clearly planned, not improvised.

Finishing 15 seconds early

Recording stops when the timer ends, not when you stop speaking. A 45-second answer on a 60-second task leaves a quarter of the response window blank. Add detail, an example, or a fuller close. Fill the time.

Wrong-topic syndrome

The prompt says "give advice." You start telling your own similar story. Beautiful English, wrong task. Read the prompt twice during prep. Write the prompt's verb on your scratch paper.

Memorised templates

Raters can tell when an answer is rehearsed. The intonation flattens, and the same connector words show up in the same places every response. Practise structures, not scripted answers.

Whisper voice

The test centre is a shared room. Keyboards click and neighbours murmur in nearby booths. Speak at a clear, projecting volume. If you'd be audible to someone two metres away, you're at the right level.

Apologising mid-response

You'll make grammar mistakes. A quick correction ("in the photo, I mean in the picture...") works. A long apology ("sorry, my English is not very good, let me start again") burns time and pulls the listenability score down by drawing attention to the slip.

Ignoring the listener

Many tasks involve speaking to a specific person: a friend, a colleague, a stranger over the phone. Address them. Use "you" instead of "a person" or "someone." Task awareness is part of the score.

Daily (15-20 minutes)

Pick one task type. Set the prep and speaking timer. Record yourself. Listen back once. Note one thing to change tomorrow. That's the whole loop.

Skip self-criticism. The point of listening back isn't to feel bad. It's to find one pattern (filler word habit, rushing in the close, missing tense consistency) and target it on the next try.

Weekly (one full attempt)

Run all 8 tasks back to back with full timing, no breaks, no restarts. This is the only way to build the stamina the real test demands. Roughly 20 minutes of focused speaking.

The 4-week curve

Keep exploring: Speaking Samples & Practice Topics · CLB Converter · CELPIP Test Format Guide · What Is a Good CELPIP Score? · CELPIP Test Day Checklist · Canadian English Vocabulary for CELPIP