1. Pick an examiner and press Start
The browser asks for your microphone. The examiner greets you and asks the first question — you answer by speaking, nothing to type.

One focused flow from choosing a test to receiving feedback — no cluttered dashboard, no guessing where your speaking stands.
The browser asks for your microphone. The examiner greets you and asks the first question — you answer by speaking, nothing to type.
The examiner moves between parts, repeats a question if you stay silent, and shows the Part 2 cue card with one minute to prepare and up to two minutes to speak.
Your test is saved in your browser first. One-tap sign-in, then the system listens back to every answer and returns a report in about a minute.
Short questions about yourself and familiar topics: where you live, work, hobbies.
One cue card. Sixty seconds to prepare, then one to two minutes of continuous speaking.
A deeper discussion around the Part 2 topic: comparisons, causes, predictions.
Fluency & coherence, Lexical resource, Grammar, Pronunciation — one band each, with a note on why it is not the next half-band yet.
Speaking rate, pauses inside clauses, fillers and over-long sentences are measured from a word-timestamped transcript. Pronunciation is heard from the original audio, not read from text.
The mean of the four bands, rounded to the nearest half band — the AI is never asked for an overall number, because that number could not be checked.
Every comment quotes your transcript verbatim. Any quote that is not in the transcript is dropped before you see it.
Grammar slips with a fix, words that were unclear, dropped final sounds.
The report ends with a single priority for your next attempt, and a model answer for Part 2.
Scores here are estimates from an automated system, not exam results. We are comparing them against recordings with official bands and will publish the measured gap in this section.
Think a score is off? The report has a feedback button pre-filled with your bands and the rubric version — that is calibration data.
During the test your voice goes to the AI examiner server to detect turns. Your answers are recorded in your browser.
Recordings are sent for marking after you sign in. The retention and deletion policy will be published before public launch.
Yes. Both the mock test and the marking are free; there is no paid tier.
Not to take the test. Marking needs a one-tap Google sign-in, because each marking run has a real cost and an account is how we cap the number of runs.
The score is an estimate from an automated system. We are comparing it against recordings with official bands and will publish the measured gap.
The four IELTS Speaking criteria: Fluency & coherence, Lexical resource, Grammar, Pronunciation. The overall band is the mean of the four, rounded to a half band, computed by code.
About 11 to 14 minutes, all three parts, like the real test.
Yes, on Chrome and Safari, once you allow the microphone. Use headphones so the microphone does not pick up the examiner from the speaker.