ScoreQwik Research
English Writing Benchmark 2026
A cross-exam view of 11,204 unique learner practice responses from Cambridge, IELTS, and PTE writing checkers.
Headline numbers
- Completed legacy evaluations reviewed
- 15,026
- Unique responses after cleaning
- 11,204
- Exam families represented
- 3
Cambridge, IELTS, and PTE
Executive summary
What this benchmark measures
The source systems recorded 15,026 completed writing evaluations. After site and exam checks, plausible-score filters, minimum-length rules, and exact-text deduplication, 11,204 unique practice responses remained: 9,158 Cambridge, 1,941 IELTS, and 105 PTE responses.
The dataset comes from the established checker applications, not from the newer ScoreQwik product database. ScoreQwik's own database currently holds 230 writing submissions and 224 completed writing evaluations. Its speaking data covers seven users and is excluded from this writing benchmark.
The report provides a shared language for comparing broad writing skills without pretending that different exams use interchangeable rubrics. Each exam-specific report remains the source for findings within that exam.
Corpus composition
Cambridge supplies most of the evidence
The bar length is each exam's share of the 11,204-response cleaned corpus. Completed-evaluation counts are shown to make the effect of cleaning visible.
Cambridge English
Benchmark cohort11,569 completed evaluations reviewed
9,158 unique responses · 81.7%
IELTS
Benchmark cohort3,222 completed evaluations reviewed
1,941 unique responses · 17.3%
PTE
Exploratory cohort235 completed evaluations reviewed
105 unique responses · 0.9%
The PTE cohort is too small for a robust standalone benchmark.
Cross-exam framework
A careful map of related writing skills
The five common domains below are an editorial framework for navigating the reports. They are not score conversions, and labels in the same row are not equivalent. Cambridge Language, for example, covers both grammar and vocabulary; IELTS distributes audience and form across task-specific criteria.
| Common domain | IELTS | Cambridge | PTE |
|---|---|---|---|
| Task fulfilment Does the response do what the task asks? | Task Achievement or Task Response | Content | Content |
| Organization and development Are ideas arranged and developed clearly? | Coherence and Cohesion | Organisation | Development |
| Grammar and language control How accurately and flexibly does the writer control language? | Grammatical Range and Accuracy | Language (partly) | Grammar |
| Vocabulary How precise, varied, and appropriate is the word choice? | Lexical Resource | Language (partly) | Vocabulary |
| Audience, purpose, and form Does the response fit the expected reader and format? | Reflected across task criteria | Communicative Achievement | Form |
What can be said now
One replicated within-exam signal, not a league table
In the IELTS Academic Task 1 and Task 2 cohorts, Grammatical Range and Accuracy had the lowest mean stored criterion estimate.
This result appeared in both large IELTS Academic task cohorts after deduplication and valid-score filtering. It describes historical practice estimates, not official examiner bands. This synthesis does not compare the numerical level of IELTS Grammar with Cambridge Language or PTE Grammar, because the rubrics and scoring scales differ.
Why PTE remains exploratory
The cleaned PTE cohort contains 105 responses. It is included so the corpus inventory is complete, but it is not large enough for the same weight of claims as Cambridge or IELTS. ScoreQwik will treat it as a pilot until the sample grows and the scoring pipeline is frozen for research use.
Methodology
How 15,026 evaluations became 11,204 responses
01
Select completed records
Only completed writing evaluations from the three legacy checker datasets entered the first-stage corpus.
02
Apply corpus checks
Records had to match the intended site and exam, contain a plausible score for that exam, and meet a minimum response length.
03
Remove exact repeats
Identical response text was counted once. Lightly edited versions may still remain because this release did not use semantic deduplication.
04
Publish aggregates
The public report contains cohort counts and high-level findings. It contains no essay text, email address, or account identifier.
Read before citing
Limitations
The users of online practice checkers are self-selected. The corpus does not represent all exam candidates, countries, first languages, or test centres. One learner may contribute multiple unique responses, so observations are not fully independent.
Historical rows do not consistently identify the evaluator model or prompt version. A change in stored estimates over time could reflect the learner cohort, the evaluation pipeline, or both. This release therefore avoids trend and causality claims.
The common-skill map supports navigation and research design. It does not convert scores across exams. All scores discussed are practice estimates; the corpus has not been validated as a substitute for official examiner scoring.
ScoreQwik is the publisher of the cross-exam synthesis. The large sample comes from its established checker datasets, not from the current ScoreQwik database. Speaking records are outside the scope of this report.
Citation
How to cite this report
Weaver, Lucas. “English Writing Benchmark 2026.” ScoreQwik, 21 August 2026. https://scoreqwik.com/research/english-writing-benchmark-2026
ScoreQwik is an independent practice platform. It is not affiliated with, approved by, or endorsed by Cambridge English, IELTS, PTE, or their respective owners. Exam names and trademarks belong to their respective owners.
Explore the criteria
Use the right rubric for your exam
ScoreQwik keeps the assessment experience exam-specific. Choose your exam to practice with its own task formats and feedback criteria.