CELPIP writing & speaking scores: what real practice answers get
Every writing and speaking answer submitted on Celpip.me is scored by AI on the official CLB scale, criterion by criterion. This page aggregates those scores — no names, no answers, just the numbers — and refreshes itself.
Last updated · 1,936 scored answers from 269 learners since July 26, 2026 · How to cite · Method at the bottom
Key findings
- 71% of 574 AI-scored CELPIP writing answers reach CLB 7 or higher; the median is CLB 7.
- Speaking scores lower: 45% of 1,362 speaking answers reach CLB 7 or higher (median CLB 6).
- The weakest criterion is Readability in writing (average 6.43) and Vocabulary in speaking (average 5.91).
- Task 4 — Making predictions is the lowest-scoring speaking task at an average CLB 5.85, against 6.62 for Task 6 — Difficult situation.
- Writing answers scored CLB 7–10 average 181 words; answers scored CLB 5–6 average 175.
Counted from 1,936 scored answers by 269 learners, July 26, 2026 – October 13, 2026. Practice scores from an AI marker, not official CELPIP results.
Writing scores by CLB level
574 scored answers. The most common result is CLB 7; 71% score 7 or higher — the level Express Entry candidates most often need.
Speaking scores by CLB level
1362 scored answers. The most common result is CLB 6; 45% score 7 or higher.
Where writing answers lose points
Average score per CELPIP criterion across 574 answers. The weakest criterion is Readability (6.43); the strongest is Task fulfillment (7.12).
Where speaking answers lose points
Average score per criterion across 1362 answers. The weakest criterion is Vocabulary (5.91); the strongest is Content & coherence (6.18).
Which tasks score lowest
In speaking, Task 4 — Making predictions averages 5.85 — the lowest of the eight — while Task 6 — Difficult situation averages 6.62. In writing, Task 1 — Email averages 7.01 against 7.07 for Task 2 — Survey response.
| Task | Scored answers | Average CLB |
|---|---|---|
| Writing · Task 1 — Email | 335 | 7.01 |
| Writing · Task 2 — Survey response | 239 | 7.07 |
| Speaking · Task 1 — Giving advice | 386 | 6.35 |
| Speaking · Task 2 — Personal experience | 191 | 6.23 |
| Speaking · Task 3 — Describing a scene | 157 | 6.51 |
| Speaking · Task 4 — Making predictions | 167 | 5.85 |
| Speaking · Task 5 — Comparing and persuading | 114 | 6.21 |
| Speaking · Task 6 — Difficult situation | 106 | 6.62 |
| Speaking · Task 7 — Expressing opinions | 131 | 6.25 |
| Speaking · Task 8 — Unusual situation | 110 | 6.19 |
Writing: practice results beside the official 2024 test
In Paragon's 2024 data report the mean CELPIP-General writing score was 7.91 (SD 1.93); 80.4% of test takers scored CLB 7 or higher and 29.8% scored 9 or higher. Our practice answers average 7.04 (median 7, 574 answers): 0.87 below the official mean. Practice answers are unrehearsed first attempts scored by an AI marker calibrated to the official descriptors — expect them to run below a paid sitting, not above it.
| CLB band | Official 2024 | Our practice | Official descriptor |
|---|---|---|---|
| 12 | 6.2% | 0.0% | Advanced proficiency in workplace and community contexts |
| 11 | 5.4% | 0.0% | Advanced proficiency in workplace and community contexts |
| 10 | 8.0% | 2.4% | Highly effective proficiency in workplace and community contexts |
| 9 | 10.2% | 10.8% | Effective proficiency in workplace and community contexts |
| 8 | 27.3% | 20.7% | Good proficiency in workplace and community contexts |
| 7 | 23.3% | 37.5% | Adequate proficiency in workplace and community contexts |
| 6 | 10.9% | 17.9% | Developing proficiency in workplace and community contexts |
| 5 | 5.3% | 6.8% | Acquiring proficiency in workplace and community contexts |
| 4 | 2.3% | 3.1% | Adequate proficiency for daily life activities |
| 3 | 0.8% | 0.3% | Some proficiency in limited contexts of personal relevance |
| 0–2 | 0.3% | 0.3% |
Official figures: Paragon Testing Enterprises, 2024 CELPIP Data Report (Tables 2, 4 and 5a); 96% of 2024 test takers took the test for immigration or citizenship. CELPIP level equals CLB level from 3 upward (Table 2). IRCC minimums link to the canada.ca page each was read from. Our figures refresh automatically; see the score chart for the full CELPIP–CLB–IELTS table.
Speaking: practice results beside the official 2024 test
In Paragon's 2024 data report the mean CELPIP-General speaking score was 7.76 (SD 2.08); 72.6% of test takers scored CLB 7 or higher and 32.6% scored 9 or higher. Our practice answers average 6.28 (median 6, 1362 answers): 1.48 below the official mean. Practice answers are unrehearsed first attempts scored by an AI marker calibrated to the official descriptors — expect them to run below a paid sitting, not above it.
| CLB band | Official 2024 | Our practice | Official descriptor |
|---|---|---|---|
| 12 | 4.5% | 0.0% | Advanced proficiency in workplace and community contexts |
| 11 | 7.9% | 0.0% | Advanced proficiency in workplace and community contexts |
| 10 | 7.1% | 0.0% | Highly effective proficiency in workplace and community contexts |
| 9 | 13.1% | 1.3% | Effective proficiency in workplace and community contexts |
| 8 | 22.5% | 13.1% | Good proficiency in workplace and community contexts |
| 7 | 17.5% | 30.6% | Adequate proficiency in workplace and community contexts |
| 6 | 14.5% | 33.4% | Developing proficiency in workplace and community contexts |
| 5 | 7.4% | 15.6% | Acquiring proficiency in workplace and community contexts |
| 4 | 3.5% | 3.7% | Adequate proficiency for daily life activities |
| 3 | 1.4% | 0.7% | Some proficiency in limited contexts of personal relevance |
| 0–2 | 0.5% | 1.5% |
Official figures: Paragon Testing Enterprises, 2024 CELPIP Data Report (Tables 2, 4 and 5a); 96% of 2024 test takers took the test for immigration or citizenship. CELPIP level equals CLB level from 3 upward (Table 2). IRCC minimums link to the canada.ca page each was read from. Our figures refresh automatically; see the score chart for the full CELPIP–CLB–IELTS table.
Longer emails do not score higher
Average length of writing answers by the CLB level they earned. Answers scored CLB 7–10 average 181 words; answers scored CLB 5–6 average 175 words. The official range is 150–200 words — the higher scores sit inside it, the lower ones run over it.
| CLB earned | Average words | Answers |
|---|---|---|
| CLB 4 | 125 | 18 |
| CLB 5 | 171 | 39 |
| CLB 6 | 177 | 103 |
| CLB 7 | 180 | 215 |
| CLB 8 | 182 | 119 |
| CLB 9 | 180 | 62 |
| CLB 10 | 189 | 14 |
Method
- Source: every writing and speaking answer submitted in practice and mock exams on Celpip.me and scored by our AI marker on the official CELPIP criteria, reported on the CLB scale (1–12). Staff accounts and our own test accounts are excluded.
- Window and sample: July 26, 2026 to October 13, 2026 — 1,936 scored answers (574 writing, 1,362 speaking) from 269 learners. The window starts on the first answer from a learner rather than a member of staff.
- Writing answers under 20 words (blank or abandoned attempts) are excluded from every figure; speaking is scored from the recorded response.
- These are practice scores from an AI marker, not official CELPIP results; the AI is calibrated to the official level descriptors but is not Paragon Testing Enterprises. Use the patterns, not the decimals.
- Percentages are rounded; per-CLB word averages are shown only where at least five answers exist. The page refreshes automatically; the raw aggregate is at /api/public/score-stats (JSON, CC BY 4.0 — cite “Celpip.me score data”).
- The people here chose to practise online, so they are not a random sample of CELPIP test takers.
How to cite this data
Please do. The figures are published under CC BY 4.0, which asks only for attribution and a link. A line you can paste:
Celpip.me (2026). CELPIP writing & speaking scores: what real practice answers get [Data set]. 1,936 AI-scored answers from 269 learners, July 26, 2026 – October 13, 2026. https://celpip.me/celpip-scores-data
- URL
- https://celpip.me/celpip-scores-data
- Last updated
- October 13, 2026
- Licence
- CC BY 4.0 — credit “Celpip.me”
- Raw data
- https://celpip.me/api/public/score-stats (JSON, same licence)
Writing something and need a figure cut a different way? Ask at support@celpip.me — we would rather you had the real number than an estimate.