How Fluensay uses AI

Last updated: 14 October 2026

Fluensay uses artificial intelligence to help you practise. Here is exactly where, what it can and cannot do, and which rules we follow (EU AI Act).

Where AI is used

• Speech recognition and pronunciation scoring — Azure AI Speech (Microsoft) turns your recording into text and measures how accurately, fluently and completely you pronounced each word and sound. The recording is not stored.
• Conversation partners — the characters you talk to in Chat are played by an AI language model (OpenAI). You are always talking to an AI, never to a person.
• Coaching feedback and weekly reports — written by the same AI model from your scores and what was recognised.
• Voices — the voices you hear are generated by Azure AI Speech; they are synthetic and do not imitate real people.

What AI does not do here

• It does not recognise or analyse your emotions, and it does not identify you by your voice.
• It makes no decisions about you with legal or similar effects; scores are practice aids, not an official assessment or certificate.
• Your data is not used to train AI models.

Limits

AI can make mistakes: speech recognition can mishear (especially with background noise), and the feedback can occasionally be wrong or imprecise. If something seems off, trust your teacher, a dictionary or your own judgement, and tell us at support@fluensay.com — a person reads every message.

The rules we follow

Under the EU AI Act, Fluensay is a consumer learning app and not a "high-risk" system (that category covers AI used by schools, universities or training providers to admit, grade or place people). We meet the transparency duties: we tell you when you are talking to an AI, and we explain here how AI is used. The AI models come from Microsoft and OpenAI, which carry the obligations for the models themselves. How your data is handled: privacy policy.