Cluster D

CELPIP Speaking: The Complete Guide to a 9+ (2026)

Most CELPIP Speaking prep fails for a predictable reason: people practice the wrong thing. They memorize templates and vocabulary lists when the test measures whether you can sound like a fluent person thinking out loud. This guide covers all 8 speaking tasks, the four criteria raters actually score, and a practice system built for the specific band you're stuck at.

Before you read another word — record yourself answering "Tell me about your favorite place in the city you live in" for 60 seconds, right now. Then take our free mock test to see how that answer scores → celpipuni.com/free-mock-test.

TL;DR

CELPIP Speaking is 8 short recorded tasks in about 20 minutes: advice, personal experience, describing a scene, predicting, comparing, a difficult situation, an opinion, and an unusual situation. Raters score content, coherence, vocabulary, and listenability — your accent does not get scored. Band 7 speakers are accurate but flat; band 9 speakers sound deliberate and natural. The fastest fix for most stuck-at-7 candidates is recording themselves daily and listening back, not more templates.

The Four Criteria (What Raters Actually Score)

The CELPIP Speaking rater has a rubric, and it's not a mystery. Four dimensions, weighted roughly equally:

  • Content/Coherence: Do you address the actual task, develop ideas with reasons and detail, and hold a logical thread from sentence to sentence?
  • Vocabulary: Range and precision — using the right word, not the fanciest one. "The queue was dragging on" beats "the line was very long" beats "the line exhibited considerable length."
  • Listenability: How easily a listener follows you — pace, pausing, stress, self-corrections that don't derail you.
  • Task Fulfillment: Did you do the job the task set out — advise, describe, compare, persuade — in the time given?

Notice what's missing: accent. A rater scoring accent would need a separate rubric, and CELPIP doesn't have one. Intelligibility is scored (they need to understand you); beauty of accent is not. This matters because the single most common wasted-effort category we see is accent-pronunciation work done by people whose actual problem is coherence.

All 8 Tasks, and What Each One Wants

The Speaking section runs about 20 minutes total. Preparation time is short (30–60 seconds per task) and response time varies from 60 to 90 seconds. Here's each task and its core move:

  1. Task 1 — Advice (~90 sec): A friend has a problem; give practical advice. The core move: name the problem, then give 2–3 pieces of advice with reasons. Don't list five ideas with no reasons — depth beats breadth.
  2. Task 2 — Personal Experience (~60 sec): "Tell me about a time you…" Pick a story you can narrate in a simple past + past continuous mix. Prep time is where you choose the story, not rehearse sentences.
  3. Task 3 — Describe a Scene (~60 sec): A photo appears. Describe what's happening. The trap: narrating a static list ("There is a man. There is a table."). The move: interpret — "the boy's mid-laugh, probably because the dog just did something ridiculous."
  4. Task 4 — Predict (~60 sec): What happens next in the picture? Speculate with probability language: "he's probably about to…", "I'd guess this is heading toward…"
  5. Task 5 — Compare (~60 sec): Two options, choose one and justify. Structure: brief take, 2–3 reasons, acknowledge the other side once.
  6. Task 6 — Difficult Situation (~60 sec): Persuade your way through an awkward spot — a friend cancels on moving day, a landlord ignores a repair. The move: acknowledge their position, ask clearly, offer something back.
  7. Task 7 — Opinion (~90 sec): "Do you agree…?" Position, 2 reasons, one concession, restate. The longest response in the section, and where template-heavy speakers sound most robotic.
  8. Task 8 — Unusual Situation (~60 sec): You see something odd and must describe it to someone who can't see it, as if on the phone. Vocabulary flexibility under mild pressure is the point.

(That's the standard structure — verify the current official task list and timings on the CELPIP site before test day, since formats get revised.)

The Band Difference (7 vs 9, Annotated)

A student of ours answered the same "describe a busy street market" prompt at band 7, then band 9, ten weeks apart. Same vocabulary range, roughly. The difference:

Band 7: "The picture shows a busy street market. There are many people shopping. Some vendors are selling fruits and vegetables. There is a woman who is buying tomatoes. The weather seems nice and sunny. Everyone looks busy but happy. In the background, there are some buildings and more people walking."

Accurate. Complete. Flat as a laminated sheet. Every sentence follows subject-verb-object. No interpretation, no stance, no rhythm — nothing a listener would remember ten minutes later.

Band 9: "This looks like a Saturday-morning market at its busiest — vendors have clearly been set up since dawn, because the produce still looks fresh and the crowd's only building. Front and center, there's a woman hand-picking tomatoes, and she's not rushing; she's doing that careful squeeze-test shoppers do when they actually care about dinner. Behind her, a vendor's leaning on his stall chatting with a regular, which tells me this is a neighborhood market, not a tourist one. Everything about it says routine — these people know each other."

What changed: interpretation over inventory ("which tells me"), narrative tenses for scene-setting, stress and rhythm, a point of view. Not one fancy word. That's the whole 8-to-9 gap compressed into one paragraph.

Mistakes That Keep You at 7

Speaking in lists. "Firstly… secondly… finally…" — that's Task 1's structure applied to every task. It's rater-repellent by Task 3.

Prepping sentences instead of ideas during the 30 seconds. Rehearsed sentences fall apart under delivery anyway. Use prep to choose your stance and your two reasons.

Filling silence with fillers. "Umm, so, yeah, I think, umm." Fillers are normal — a few per minute are fine and even human. A filler every fourth word reads as a vocabulary retrieval problem.

Stopping at your first idea. Most candidates' first idea is their shallowest. The 9-speaker habit: take the second idea, the one that came with a "wait, actually —"

Practicing silently. You cannot improve speaking by reading about speaking. This sounds obvious, and it is the #1 reason people plateau: their practice never opens their mouth.

A Practice System That Actually Moves Bands

The system we use with students, daily, 25 minutes:

  1. Record (10 min): Two tasks from different types, timed, out loud, standing up. Standing changes your breath, and breath changes your voice.
  2. Listen back with the rubric (5 min): Score yourself on the four criteria, 1–12 each. Be brutal — you are not the audience you're trying to convince.
  3. Redo the weaker one (5 min): Same task, once, fixing the single weakest criterion. Not all four. One.
  4. Shadow native audio (5 min): Play 30 seconds of a podcast and re-speak it in real time, matching rhythm and stress. This is where listenability comes from — not from studying it, from copying it.

Week over week, that's 14 recorded tasks and 7 shadowing sessions. Students who follow it for six weeks move, on average, close to a full band. Students who do the same tasks in their heads move zero.

If you want a trained grader to find the gap you can't hear yourself, our speaking evaluation pinpoints exactly which criterion is costing you the level → celpipuni.com/evaluation.

The One Nuance Nobody Mentions

Here's something the template-sellers won't tell you: your Task 1 and Task 8 performances matter more than their point values suggest, because they're the rater's first and last impression, and raters are human. A strong Task 1 sets an anchor; a strong Task 8 leaves one. When energy is limited — and on test day, it is — protect the openers and closers, and let the middle tasks be "very good" instead of perfect.

The counterintuitive implication: if you're going to use any template anywhere, use it in the middle tasks, where the rater has already formed an impression, and go free-form on the bookends.

Your Checklist for This Week

  • Record two tasks daily (any type), timed, out loud
  • Listen back once with the four criteria and score yourself
  • Redo the weaker task once, fixing one criterion
  • Shadow 5 minutes of native audio daily
  • Take one full 8-task practice run under real conditions before test day

FAQ

How long is CELPIP Speaking?

About 20 minutes, embedded in the ~3-hour test. You'll speak 8 responses of 60–90 seconds each, with 30–60 seconds of prep between tasks.

Is accent scored in CELPIP Speaking?

No. Intelligibility is scored — the rater must be able to follow you. Accent itself is not a criterion. Clear pronunciation helps your listenability score; sounding Canadian is not required.

Can I retake just Speaking?

No. CELPIP is a single certificate covering all four skills; improving one module means retaking the full test.

What if I go blank mid-task?

Self-correct and pivot: "Actually, the better example is…" Raters expect imperfection under time pressure; how you recover shows real fluency. A smooth recovery scores better than a memorized answer delivered shakily.

Are templates bad?

Reusing structures is fine and smart. Reusing sentences across tasks is what reads as band-limiting. If your template survives swapping the topic, it's too rigid.

Do I speak to a computer the whole time?

Yes — into a headset at your station, recorded for rating later. There's no examiner in the room with you, and other test-takers around you will also be speaking. Practicing with mild background noise is genuine test prep.

What to Do Next

  1. Record your first two tasks today — before you feel ready (readiness is a feeling, not a prerequisite).
  2. Score yourself against the four criteria tonight.
  3. Book your real test date now if you haven't — a deadline converts practice into performance.
  4. Get one professional evaluation this week to calibrate your self-scores.

The Bottom Line

CELPIP Speaking rewards people who sound like they're thinking out loud in fluent English — not people who recite memorized English at a microphone. The tasks are predictable, the criteria are public, and the practice loop (record → listen → fix one thing → shadow) works if you actually open your mouth daily.

Find out which criterion is holding your score back — get your speaking evaluated by a trained grader → Get Your Speaking Evaluated.