CELPIP Speaking Tips
Tips taken from the four official criteria rather than from folklore, one for every task in the 2026 format. Then test them where they are hardest to apply: Test 6, the toughest of the six here.
Most CELPIP speaking tips circulating online are folklore: talk fast, use big words, always argue about price. Some of it is harmless, some of it costs marks. The tips below come from what the section actually measures, and from one fact that reorganises everything else.
That fact is this: given the same picture, the CLB 5 candidate and the CLB 12 candidate notice the same things. The gap is not observation, it is the descriptive machinery available to render what they saw. So tips about spotting more detail are aimed at the wrong target. Precision, hedging, complex clauses and tuning your words to your listener are the target.
Four CELPIP speaking tips that apply to every task
Listenability names pauses, interjections and self-correction directly. A clumsy sentence that arrives costs less than a clean one you restarted twice. Finish the thought and move on. Silence scores nothing at all.
Running out of things to say at 40 seconds is a symptom of listing. Content rewards the quality of ideas and the detail supporting them, so one idea with a reason, an example and a consequence beats three bare assertions and fills the window without padding.
Vocabulary is scored on suitability and precision, not rarity. "The registration table" is worth more than "the thing over there", and a rare word used slightly wrong costs you twice, once on accuracy and once on the pause you took to retrieve it.
Task Fulfillment marks relevance, completeness, tone and length. Most lost marks here are a missing final beat: the request never made, the other option never addressed, the person you were speaking to never really addressed as a person.
A tip for each of the 8 tasks, using Test 6
Test 6 is the hardest test on this site, so it is the right place to check whether a tip is real. Each tip below is aimed at that task's own subject on this test.
Task 1, Giving Advice (30s prep, 90s speak). Your friend's teenage son has stopped answering her messages. Advice tasks collapse into platitudes because candidates advise the topic instead of the person. Recommend something she can do on Tuesday, give the reason, and say what it costs her. Ninety seconds of "communication is important" is a long way to say nothing.
Task 2, Personal Experience (30s prep, 60s speak). A time you gave someone news they did not want. One story. Candidates lose this task by summarising three. Sixty seconds is enough for a situation, a turn and a consequence, and nothing else.
Task 3, Describing a Scene (30s prep, 60s speak). A hotel lobby during a conference: a clerk handing a lanyard across the registration table, a concierge pointing down a hallway, two porters and a green suitcase sliding off a luggage cart, a server refilling a coffee urn, a woman holding a full cup in each hand and leaning at a tray of muffins she has no hand free to take, a lanyard check at the doorway. Build a frame before you build detail: left, centre, right, back. Your listener cannot see it, so a random tour of the room is unusable to them.
Task 4, Making Predictions (30s prep, 60s speak). Same lobby, no new picture. Predict from what is mid-motion: the sliding suitcase, the woman with no free hand. Then hedge properly. "It will fall" is a weaker answer than "it looks like it is about to slide off, so the second porter will probably catch it before it hits the floor", because the hedging is the grammar being marked.
Task 5, Comparing and Persuading (60s to choose, then 60s prep, 60s speak). Your walking club's two-day trip: the Ridge Trail, 22 kilometres with a three-hour rocky climb, open valley views, carry your own tents and water; or the River Valley Loop, flat gravel, cabins with bunks at the halfway point, four waterfalls, phone coverage the whole way. Then your friend counters with a guided canoe route where everything is provided and nobody navigates, but every paddler must pass a swim test and wind cancels it. Note what is missing: there are no prices anywhere on this task. The popular tip to "always argue cost first" leaves you mute here. Argue on fit for the group, on risk, on what happens if it rains, on who gets left behind.
Task 6, Difficult Situation (60s prep, 60s speak). You build the schedule. One evening shift is left. Dana has worked mornings only for two months because she cares for her mother in the evenings; Marc has worked every evening for those two months and has already asked you to stop. You must tell one of them. Both are right, so whichever you choose, name the other person's fairness out loud before you deliver the decision, and deliver it plainly. Tone is scored, and evasion reads worse than the bad news.
Task 7, Expressing Opinions (30s prep, 90s speak). Should there be a limit on how much more the highest-paid person in a company can earn than the lowest paid? Take a side in the first sentence. Ninety seconds of even-handed hedging is not fairness, it is thin content, and it leaves you nothing to develop.
Task 8, Unusual Situation (30s prep, 60s speak). A rusty green hand-cranked machine at a flea market: a brass funnel on top like a watering can mouth, a crank whose wooden knob is split with one half hanging, three brass rollers behind an open panel with a chunk missing from the middle one, a spring dangling unhooked, a dented chute, a snapped lever, and a missing clamp screw so the base tilts. It has no name and you cannot photograph it. Anchor the size first, work part by part, compare to things Priya knows, and finish by asking her whether it can be repaired. That closing request is scored, and it is the beat candidates most often drop.
Tips to ignore
- "Speak fast to fit more in." Rhythm and intonation are marked. Rushing damages both and rarely adds an idea worth having.
- "Memorise an answer for each task type." A recited paragraph fits the question you were given only by accident, and recited speech is audible.
- "Always argue about price on Task 5." Test 6's Task 5 has no prices. Official items behave the same way. The skill is comparing on whatever dimensions exist.
- "Use as many idioms as possible." Suitability is scored. An idiom that misses register costs more than the plain word it replaced.
- "Task 8 is always a weird object." It is whatever has no ready-made label, which sometimes means a broken machine and sometimes a situation you are stuck in. Prepare the skill, not the guess.
Where the tips get tested
Test 6 runs all 8 tasks on the official 2026 clock: 30 seconds of preparation and 90 of speaking on Tasks 1 and 7, 30 and 60 on Tasks 2, 3, 4 and 8, and 60 and 60 on Tasks 5 and 6, plus Task 5's silent 60-second choosing screen. About 15 minutes in total, the timer resetting at each task and preparation never skippable. Every task is scored on Content and Coherence, Vocabulary, Listenability and Task Fulfillment, and the report returns expert feedback per criterion with an estimated CLB level from 1 to 12.
If you are working through the set, this is the one to sit last. Reading a tip costs nothing. Applying it at 60 seconds a task, with no prices to argue about and two colleagues who are both in the right, is the part that shows you whether you have it.
Free, the hardest of the six, with expert feedback and a CLB estimate.