Methodology

Current data: ChatGPT via API with web search, results may differ from the app. Other assistants: scan coming soon.

1. For each suburb and assistant we send these five prompts, each three times, with a Sydney location hint:

   Who are the best dentists in {suburb}, Sydney?
   I need a dentist in {suburb} NSW. Which practices would you recommend?
   Emergency dentist near {suburb} Sydney open today?
   Affordable family dentist in {suburb}?
   Which dental clinic in {suburb} is good for nervous patients?

2. Assistants are queried by API with web search on: ChatGPT (OpenAI Responses API with web_search); Gemini (Gemini API with Google Search grounding); Perplexity (Perplexity Sonar API); Grok (xAI API with live search).
3. We extract practice names from each answer and match them to a list of practices in the suburb.
4. Score = share of (prompt × run × assistant) answers that named the practice. Per-assistant scores use only that assistant's answers.
5. We compare with the previous scan to show movers. Excerpts are quoted verbatim with the date and prompt; mentions of star ratings or reviews are omitted and marked […] (AHPRA testimonial caution).

Limits

• API answers can differ from the consumer apps, and vary by run, account, model version and location.
• Assistants can make mistakes, including naming closed or wrong practices. We quote them; we don't endorse what they say.
• A score is not a measure of clinical quality.