Instant.
Accurate.
Reliable.

What is the Dynamic Speaking Test?

The Dynamic Speaking Test is a human-designed, AI-graded assessment that produces a detailed CEFR-based report on a test taker’s speaking ability. It takes around 30 minutes to complete. Its scoring has been calibrated against a panel of experienced examiners, so results are consistent with what a human expert would award.

If you can’t see the video, click here.

Time stamps and transcripts →

  • 0:00 What is the Dynamic Speaking Test?
  • 1:00 Key features: Accuracy / AI marking
  • 1:30 Key features: Validity
  • 2:00 Key features: Efficiency
  • 2:30 The test: an overview
  • 3:02 Assessment areas
  • 3:35 How does it work? / Test demo
  • 4:35 Administration overview
  • 5:10 Support

Why choose the Dynamic Speaking Test?

Runs on laptops, tablets or phones

Convenient

The test runs on laptops, tablets or phones, and takes around 30 minutes. Results come back automatically, so you can place a whole class in one session instead of running speaking assessments one at a time.

AI grading designed by assessment specialists

Accurate and reliable

Tasks are written by experienced assessment specialists, and the AI grading produces consistent scores from one session to the next.

Detailed five sub-skill results

Detailed

Every result breaks down into five sub-skills: pronunciation, fluency, grammar, vocabulary and task achievement. Teachers can see where a learner is doing well and where they need more support.

How it works

1. The learner sits the test

Test takers work through a series of real-world speaking tasks – giving opinions, making suggestions, solving problems. Tasks are randomised and increase in difficulty as the test progresses, so two learners sitting next to each other won’t get the same test, and more able speakers get the chance to show their full range.

A learner taking the Dynamic Speaking Test with a headset

2. The AI grades the response

The AI scores each response across five sub-skills: pronunciation, fluency, grammar, vocabulary and task achievement. Crucially, it doesn’t just transcribe what the learner says – it also judges whether the response actually addresses the task, so a memorised answer won’t earn a high score.

The AI grading the five speaking sub-skills

3. You get the results

As soon as the test is over, you receive a CEFR-based report for each learner. It includes an overall level, a position within that level, and a spidergram showing the five sub-skills. Results can be exported to your school’s management system.

A Dynamic Speaking Test certificate with CEFR level and spidergram

Dynamic Speaking Test blog & news

The Unbearable Tedium of the Speaking Test

The Unbearable Tedium of the Speaking Test

Andrew Stokes · April 2026

Manual speaking tests are challenging because they are: time-consuming, expensive, difficult to grade and benchmark, and are prone to bias. What if an AI-graded test solved all those problems?

Read the post →
Why AI needs humans – and humans need AI – to deliver an effective speaking test

Why AI needs humans – and humans need AI – to deliver an effective speaking test

Martin Moore · January 2026

Can a machine match a human’s ability to assess a learner’s proficiency in English? Like all uses of AI – it depends on the right combination of human skills and machine capability. Martin Moore, Head of Testing at Clarity, explains how.

Read the post →
AI speaking assessment: How we designed a valid test

AI speaking assessment: How we designed a valid test

Martin Moore · April 2024

The road to our AI speaking test wasn’t a smooth one. Martin Moore lays out the trials, tribulations and triumphs of developing the Dynamic Speaking Test.

Read the post →