The Dynamic Speaking Test (DST) uses an AI trained against thousands of human-graded student responses. In a 2024 calibration study, the AI reached the same level of accuracy as expert human markers – i.e. its scores fell within one CEFR level of other markers in 98% of cases. Unlike human examiners, the AI is not subject to fatigue or inter-rater variability – the fact that two qualified examiners often grade the same response differently. Every student in your cohort is judged by the same standard.

The AI analyses more than 35 linguistic features simultaneously, across three categories:

• Acoustic features: pronunciation and intonation.
• Linguistic features: grammatical complexity and lexical range.
• Discourse features: coherence, flow, and relevance to the task.

These are mapped to CEFR descriptors to produce a level (A1 to C2) and a numeric position within that level.

The AI is trained on a diverse dataset of English speakers from different linguistic backgrounds, which prevents bias against particular regional pronunciations. The AI assesses intelligibility and clarity rather than accent itself.

No – there are several safeguards in place.

1. The test checks whether students actually answer the question. Unlike basic speech-to-text tools, the DST evaluates task achievement. If a student delivers a fluent but irrelevant response – reciting a memorised passage when asked to solve a problem, for instance – the AI spots the mismatch and scores accordingly. This reflects the CEFR’s focus on communicative competence: it’s not about how much English a student can produce, but whether they can use it to do the job the question asks of them.
2. Questions are unseen and randomised. Students don’t know the questions in advance, and the system draws from a large question pool, so no two tests are the same. Students sitting side by side will get different questions – copying isn’t possible.
3. Answers are checked for originality. The system compares spoken responses against a wide database of online material, so memorised or scripted answers get flagged. Students need to respond in their own words, in the moment.
4. On-screen activity is monitored. If a student clicks away from the test window to look something up or get help elsewhere, it’s logged. (Only activity within the test itself is tracked, so privacy is respected.)

Together, these measures mean the test reflects what a student can genuinely do in English – real communicative ability, not memorisation or preparation strategies.

The AI marks each response in seconds. Results are available as soon as each student finishes – there is no marking queue, regardless of cohort size. Teachers receive the full set of class results, along with individual test taker certificates, immediately after the test finishes.

The DST is designed for placement and progress monitoring within an institution, rather than as a high-stakes external test. Because the tasks are communicative and CEFR-aligned, the DST works well as an affordable progress check or as preparation for students working towards IELTS, TOEFL or Cambridge exams. Teachers and students can see where they currently stand before committing to an expensive external test.

Yes. The interface is deliberately minimal – a single screen with one record button – so students do not need prior experience with online testing platforms. Each session also begins with a hardware check and a warm-up task, so students can get comfortable with the format before the assessment begins.

Yes. All results can be exported as CSV or Excel files, ready to upload into Moodle, Canvas, or your own school system. This lets you track student progress over time without any custom IT work.

Yes. The DST and the wider ClarityEnglish platform are fully GDPR-compliant. We use strict administrative and technical safeguards to keep learner data secure, and users have the full set of rights guaranteed under European data protection law – including the right to access their data, request corrections, and restrict how it’s used.

For test administration, the platform stores basic identifying information – name, email address, and institution. Where schools prefer not to use real names or emails, a student number and a dummy email address work just as well. Alongside this, we store the student’s speaking recordings and their performance data (scores and progress records).

This information is used solely to process results and assess CEFR sub-skills. It isn’t shared with third parties or advertisers, and access is limited to authorised ClarityEnglish staff for the purpose of running and improving the platform.

Student data and test recordings are stored on secure servers in the United States and Ireland. Both locations operate under strict data protection standards, with Ireland’s servers falling under the EU’s GDPR framework – widely regarded as one of the strongest data privacy regimes in the world.