How we write our tests

1. A benchmark from Cambridge's own samples

Before any test was written, Cambridge's two official sample tests for each exam (also reproduced in the B2 First and C1 Advanced Handbooks for Teachers) were measured part by part: text and sentence lengths, what each question tests, rubric wording, and the length of the listening scripts. These measurements became the targets every new test must meet. Official texts and questions are never copied or adapted; only the measurements carry over. Format facts come from the handbooks and the B2 First and C1 Advanced exam format pages.

2. Written by our team from a detailed brief

Each test is written by our team from a detailed brief: the exact format of every part of B2 First and C1 Advanced, the benchmark ranges, and our item-writing rules: one correct answer, plausible wrong options, no specialist knowledge, no upsetting topics, and topics spread across tests. The writing stage includes its own answer check, in which every question is answered twice without the key.

3. Automatic checks

Validation scripts check every test before review: the number of parts and questions, question numbering, that every gap has a question and every question a valid key, that key word transformation answers stay within the word limit and keep the key word unchanged, that every Listening answer is said word for word in the script, that every Speaking task has its photos, and that text lengths fall inside the benchmark.

4. Expert review and a blind answer-key check

We do not rely on the writing stage's own checks. Every test is reviewed by an expert, and every Reading, Use of English and Listening question is answered again from an answer-free copy of the test (with the scripts in place of audio), without the key or the earlier checks. The review also covers ambiguity, giveaways, wording and level, and the Writing and Speaking tasks. Every item changed as a result is solved blind again.

• 1,008 objective questions solved blind across the 12 mocks (492 B2 First, 516 C1 Advanced).
• 1,007 matched the key first time. The one difference was a second correct answer, which is now accepted. No question was wrongly keyed.
• About 40 further correct answers that the key would have marked wrong were added, and items where two answers fitted, or where one question gave away another, were rewritten.
• All 30 rewritten items agreed with their new key when solved blind again.

5. Audio and images

Listening scripts are recorded with synthetic neural text-to-speech voices in a range of English accents, then assembled with the rubric, the reading times in Cambridge's sample recordings (estimated where Cambridge gives none) and the second hearing. Our Listening papers run between about 39 and 42 minutes, including both hearings; the real paper takes about 40. Speaking photos are generated images of everyday scenes, checked by eye: no text, logos or real people.

6. A full play-through

Every test is run through the real scoring code with right, blank, wrong and half-right answer sheets for each full mock, paper and mini mock (2,088 automated checks), and the public test files are checked so that no answer can be read from them before you submit.

What we have not done

No learners have piloted the tests, and our scores have not been calibrated against official results. The voices are synthetic. These are original practice tests, not Cambridge papers adjusted to the same difficulty, which is why every scale score on the site is labelled as an estimate.

Corrections and changes

If you think an answer is wrong, email hello@cambridgemocktests.com or use the contact page, with the exam, the test, the paper and the question number. We check every report, correct the test if needed and record the change here. We never publish answers in this log.

Last updated 15 November 2026.