How we score — honestly

An estimate is only useful if you can trust it. Here is exactly what the Fluency + Accent Check measures, how, and where the limits are. We would rather under-promise than show you a number we can't stand behind.

What the Fluency + Accent Check measures

How the AI + Accent Audit work

The accent meter is powered by a speech-assessment engine that compares your recording to a reference and returns per-sound accuracy, fluency, and prosody scores — the same engine that gates progression inside the course, where a sound only counts as mastered after several strong attempts across different words (so you can't fake fluency by memorizing one phrase). Your AI roleplay partner gives unlimited judgment-free conversation reps; the meter, not your feelings, decides when you've leveled up.

When you should add a human

Use a live human conversation — a 30-minute session, a coaching sprint, or an interview rehearsal — when you need real, unscripted talk under social pressure: before a job interview, when you plateau, or when you want to handle the messy disagreement an AI can't fully fake. The AI loop is enough to reach B2 on its own; the human hour is the optional finish.

What we never do

We never call a Ledoly level "official", "accredited", "endorsed", or "guaranteed". We never promise a specific level by a date. And if the accent scoring is ever down or low-confidence, the system says so and refuses to pass a Conversation Challenge on a fabricated number — the honest-failure rule is hard-coded.

Your data & privacy

Your typed responses and any voice recordings are processed only to estimate your level and score your pronunciation, and your speech is treated as sensitive personal data. We tell you before any recording, and you can take the typed Fluency Check instead of the spoken one.