Skip to main content
Practicly

Understanding AI in Practicly

Last updated: 15.09.2026 Version: 1.0

1. What is the AI doing in your session?

When you start a session in Practicly, you talk to a virtual patient — a simulated person played entirely by AI.

Here is what happens, step by step:

  1. Your voice is captured by your microphone and converted to text in real time.
  2. The text goes to a language model (a type of AI). The model reads the patient's profile and generates what the patient would say next.
  3. That response is turned back into speech by a text-to-speech AI and played through a digital avatar.
  4. At the end of the session, an AI reviews the transcript and produces a written feedback report for you.

Nothing in the session involves a real person playing the patient. Every patient utterance, every avatar gesture, and every feedback comment is AI-generated.

2. Can the AI be wrong?

Yes, and it is important that you know this.

AI language models sometimes produce information that sounds confident but is factually incorrect. In a medical-training context this could mean:

  • A patient describing a drug interaction that does not actually exist.
  • Symptoms that do not match any real condition.
  • Lab values that are outside normal physiological ranges.

This is called hallucination. It happens because the AI generates text based on patterns, not because it "knows" medicine.

The session feedback is advisory only. It is not a clinical assessment, not a pass/fail grade, and it is never shared with your institution or employer.

3. When should you NOT use Practicly?

Do not use Practicly for:

  • Real patient care. Never apply what a virtual patient says to a decision about a real person.
  • Clinical diagnosis or treatment. The AI can reflect common presentations, but it cannot diagnose.
  • Board certification or licensing exams. Practicly is a practice tool, not an accredited assessment.
  • Any safety-critical decision. If a situation affects a real person's health or safety, use verified clinical resources and qualified colleagues — not this app.

4. How to report a concern

If you see an AI response that seems wrong, harmful, or unexpected, use the "Report AI response" button. You'll find it in the conversation transcript header during a session, and on the session summary screen.

Use the button rather than email — it automatically attaches the relevant part of the session (transcript and feedback) to your report, so our team can see exactly what you're referring to. A session ID alone isn't enough, because session data is protected.

We review every report. Urgent concerns are escalated within one business day.

5. Your rights

Human review. If you believe an AI-generated feedback report is unfair or inaccurate, you can request a review by a member of our team — use the "Report AI response" button on the session summary screen. Your report automatically includes the session context, so you don't need to provide a session ID manually.

Your data. Your voice and transcript are processed to run the session and generate feedback. They are not used to train AI models. Transcripts are kept for up to 24 months and then deleted. Full details are in our Privacy Policy.

GDPR rights. If you are in the EU, you have the right to access, correct, delete, or export your data. Contact [email protected].

More information. See the full AI Disclosure page for technical details about the models we use and how we govern our AI system.

This document is part of our EU AI Act Article 4 AI literacy programme and is in effect as of 15.09.2026.