Frequently asked questions

Closed beta — send anything missing to drshreydhawan@gmail.com.

1. 1. What is this tool?

   An AI patient simulator for the Australian Dental Council practical exam. You pick a station (consent, perio, paeds trauma, an emergency, etc.), the AI plays the patient, you talk or type back, and Claude Sonnet scores your performance against the station rubric using the same grade bands the ADC uses (very good / satisfactory / borderline / unsatisfactory + global rating). The same engine works for AMC, NDEB, ORE and other OSCE-style exams; ADC dental is the first content pack.

2. 2. Is this affiliated with the Australian Dental Council?

   No. SQH Technologies built and runs this independently. The ADC has not endorsed, reviewed, or licensed any content here. Use it as practice, not as a predictor of your actual exam result.

3. 3. Where do the stations come from?

   Some are hand-authored, most were AI-generated from a topic list curated by a practising dentist. Each station has a candidate brief, a private examiner context, a 5–8 item rubric, and a patient persona. We refine these continuously — flag any inaccuracy via the feedback form and we usually fix it within a day.

4. 4. Why does the patient voice sound a bit robotic?

   During closed beta we use the browser's built-in speech synthesis to keep cost at zero per attempt. On a Mac, install a premium en-AU voice (System Settings → Accessibility → Spoken Content → "Karen (Premium)") and the tool will pick it up automatically. The production version will switch to a higher-quality TTS (OpenAI's gpt-4o-mini-tts), at which point voices will sound noticeably more natural and persona-specific.

5. 5. How long is each station?

   We default to 8 minutes per station — that's the widely-cited ADC clinical-skills-day duration (with reading time at the door). The timer auto-starts when you press Start and auto-fires scoring at 0:00, exactly like the real exam bell. You can also hit "Score now" early if you're done.

6. 6. How accurate is the AI scoring?

   Reasonably accurate at the per-criterion level — Sonnet 4.6 is good at spotting whether you mentioned, e.g., dry socket as a consent risk and at quoting the line of your transcript that scored each criterion. It's less reliable on global pass/fail at the margin, especially on technical-skill stations where you're describing rather than doing. Treat the rubric report as a checklist of what you missed, not as a verdict.

7. 7. What happens to my transcripts?

   They live in our Neon Postgres database, attached to your account. We use them to render your Dashboard. We do not use them to train any AI model, and the underlying API providers (Anthropic, OpenAI) are stateless on our side. You can export everything or delete it at any time — email drshreydhawan@gmail.com and we'll do it within 7 days. Full detail on the Privacy page.

8. 8. Is my data shared with Anthropic or OpenAI?

   Yes — your transcript is sent to Anthropic to generate the patient turns and the score, and your patient lines are sent to OpenAI for TTS. Both providers state they do not retain or train on API traffic. The candidate brief and rubric are kept on our server and never leave it.

9. 9. Why do I need to sign in?

   Two reasons. (1) AI calls cost money per attempt — without auth, anyone who finds the URL could rack up our bill. (2) Your dashboard, attempts history, and progress trends are private to you. Sign-up is via Clerk (email + password or Google), takes 30 seconds.

10. 10. Can I share my score?

    Not in beta — your dashboard is private to your account. A "share my result" feature with an anonymous public URL and a downloadable image card is on the roadmap.

11. 11. What does it cost?

    Beta is free. Each attempt costs us roughly $0.04 USD in AI charges, which we cover during testing. Future pricing is likely a tiered subscription (free taster + $49/mo Prep + $99/mo Exam Sprint AUD); we'll give beta testers a discount at launch.

12. 12. I think a station rubric is wrong. What do I do?

    Email drshreydhawan@gmail.com with the station title and what's off — we'll regenerate or hand-fix it. AI-generated stations are good but not always perfect, especially around guideline-specific clinical detail (anticoagulant dosing, paeds drug doses, etc.). Flag them aggressively.

13. 13. How do I report a bug or request a feature?

    Same email. Screenshot helps. We deploy multiple times a day during beta so most fixes ship the same day.

14. 14. What stations are available right now?

    ~65 stations covering communication & ethics, perio, endo, paeds, oral surgery, oral medicine, prosthodontics, restorative, special-needs dentistry, orthodontics, and medical emergencies. Open the dropdown on Practice to see the full list grouped by category.

15. 15. Will this expand beyond dental?

    The engine is exam-agnostic — same code can run AMC clinical (medicine), NDEB (Canada), ORE (UK), and pharmacy / nursing OSCEs with different content packs. ADC dental is intentionally the first, because the founder is a dentist and the gap is acute. Other exams will follow once beta is stable.

© 2026 SQH Technologies · Not affiliated with the Australian Dental Council.