Where does AI sit between the phone call and the ad?
It sits at four steps in a patient’s path, and at each step the rule is the same: the tool may help, and a named person signs.
Take one patient. They ring or message the practice. They are seen and X-rayed. A note and a letter follow. Before any of that, they found the practice through an ad or a page.
| Step | What a tool may do | Who signs | Source and date |
|---|---|---|---|
| The phone call or chat | Give approved facts, capture a message, say it is automated, collect only what the message needs | The front desk takes emergencies and treatment questions | OAIC, updated 17 January 2025 |
| The X-ray or photo | Where the tool’s ARTG entry covers the job: highlight or aid detection | The dentist makes the diagnosis | ADA Policy 6.34, 17 November 2023; TGA list, 12 October 2026 |
| The note or letter | Draft the words | The person who sends it checks it | ADA Policy 6.34, 17 November 2023 |
| The ad or page | Draft the words | A person approves each public sentence | Ahpra guidelines, reviewed 2 September 2025 |
Swipe the table sideways to see every column.
What can it do at the front desk?
It can answer questions the practice has approved, take a message, and hand the patient to a person.
The OAIC’s guidance on commercially available AI products (published 21 October 2024, updated 17 January 2025) sets the ground rules for a chat on a practice website:
- Say plainly that it is a chatbot.
- Collect only what the message or booking needs.
- Put a plain line about the AI in the privacy policy and the collection notice.
- Hand emergencies and treatment questions to a person.
Shoutout Digital’s chat agent follows that shape: it hands enquiries to the front desk and does not book. Read how AI live chat works, or the difference between an AI agent and a chatbot.
What can it do with an X-ray or a photo?
A tool that reads dental images can be a medical device. The TGA decides that by the tool’s intended purpose, and an ARTG entry then says what the tool is for.
The TGA publishes a list of AI-enabled medical devices in the ARTG (announced 4 September 2025; the dental entries below were read on 12 October 2026). The TGA says the list may not include every device that uses AI. Where a tool is on it, the right phrase is “included in the ARTG as a medical device, for this intended purpose”. Notice what the entries say about the dentist.
| Tool and ARTG number | Class on the list | What the entry says it does | The limit the entry states |
|---|---|---|---|
| Pearl Second OpinionARTG 427104. Sponsor: PharmaDEV Consulting Pty Ltd | Class IIa | Computer-aided detection on bitewings, periapicals and panoramics. For patients aged 4 to 11, it is limited to caries. | A second reader for the clinician. |
| OverjetARTG 511525 and 511529. Sponsor: Emergo Asia Pacific Pty Ltd | Class IIa | Aids detection of periapical radiolucencies (511525) and of caries on bitewings and periapicals for patients aged 4 to 11 (511529). | Not a replacement for the dentist’s review. |
| Align X-ray InsightsARTG 489427. Sponsor: Invisalign Australia Pty Ltd | Class IIa | Aids detection on bitewings, periapicals and panoramics, for permanent teeth. | Used with the dentist’s own review; should not be the sole basis of a diagnosis. |
| Diagnocat AISARTG 481005. Sponsor: Taevas Life Sciences Pty Ltd | Class IIa | Cloud processing of dental images, for education and highlighting. | Cannot be used for direct diagnosis, or to decide that disease is present from the screening alone. |
| CoTreatARTG 400152. Sponsor: CoTreat Pty Ltd | Class 1 | Dentists upload photos and radiographs. It makes observations on caries, wear, calculus and other findings. | Empirical observations, not diagnoses. |
| SmiloARTG 335634. Sponsor: Oral Tech Al Pty Ltd | Class 1 | A consumer app. Photos of teeth and gums are scored for awareness of caries and periodontitis. | No diagnosis; the user is told to see a dentist. |
Swipe the table sideways to see every column.
The list also has two charting tools: EM2Clinic (ARTG 483608), which highlights anatomy and past treatment and is not for direct diagnosis, and Overjet Charting Assist (ARTG 512268). Source: TGA list, retrieved 12 October 2026. The table records what each entry says the tool is for.
Two other points sit on the same page of the TGA’s guidance (last updated 5 February 2026). A tool that later adds diagnosis or treatment suggestions has changed its intended purpose. The TGA calls that scope creep, and the new version needs ARTG inclusion before it is supplied. And a consumer scoring app such as Smilo is a different product from a dentist’s own diagnosis. A practice should not advertise a patient’s Smilo score as a result.
What can it draft, and what must a person check?
It can draft notes, letters, review replies and ad copy. A person must read each one before it leaves the practice.
ADA Policy 6.34 names the wider map: radiographs and photographs, charting and treatment planning, progress tracking, practice efficiency and business analytics, insurance-claim analysis, and generative drafting of notes and letters. It limits the generative uses to low-risk drafts that a person checks.
Digital scribes have their own TGA page (30 January 2026). A scribe that only records and summarises is not automatically a device. A scribe that interprets the conversation into a diagnosis or a treatment recommendation is.
Anything public also meets section 133 of the National Law, as explained in Ahpra’s advertising guidelines. The ad must not be false or misleading. It must not use a testimonial about a clinical aspect of care. It must not create an unreasonable expectation of benefit. It must state the terms of any gift or discount. The rules bind a page, a search ad and a sentence a model wrote. For the ads we run, we agree the budget with the practice, then our team and agents run the campaigns 24/7. Agents never spend past the agreed budget, and every change is logged.
What should staff keep out of ChatGPT?
Staff should keep a patient’s name, date of birth and history out of any personal or free AI account.
The OAIC says personal information put into an AI system is still personal information, and so is personal information the system generates. Its best-practice line is not to enter personal information, and particularly sensitive information, into a publicly available generative AI tool. Clinic Comply (6 October 2026) restates that for practices: no patient details in a personal or free ChatGPT or Copilot account.
What should patients be told?
Tell patients when their information goes into a tool, in plain words, and tell them what a public chat tool cannot do.
Ahpra’s news item of 22 August 2024 keeps practitioners under their code of conduct. The principles it names are accountability, an understanding of the tool, transparency, informed consent, and the ethical and legal issues. The Australian Commission on Safety and Quality in Health Care’s AI Clinical Use Guide (version 1.0, August 2025) adds that clinicians should check the evidence first, expect invented output, and keep records to the usual standard.
ADA Policy 6.34, clause 2.14, says practitioners should tell patients about the risk of asking a public generative AI system for a diagnosis or a treatment plan.
Questions
Is dental X-ray AI legal in Australia?
It can be. A tool that is a medical device must be included in the ARTG before it is supplied for its purpose, and the TGA decides that by intended purpose, not by the word “AI”.
Several imaging tools are on the TGA’s list of AI-enabled devices in the ARTG, read on 12 October 2026. Examples are Pearl Second Opinion, Overjet and Align X-ray Insights. Each entry names what the tool is for, such as aiding detection of caries on bitewings. The TGA says the list may not include every device that uses AI, so check the tool’s own ARTG entry before you buy.
Being on the register does not mean the tool decides anything. The entries say the dentist still reviews the image.
Does the dentist still have to look at the X-ray?
Yes. The dentist makes the diagnosis, using the history, the examination and the tests. That is ADA Policy 6.34, amended 17 November 2023.
The ARTG entries say the same in their own words. Overjet’s entries say the tool is not a replacement for the dentist’s review. Align X-ray Insights says its results should not be the sole basis of a diagnosis. Diagnocat AIS says it cannot be used for direct diagnosis.
Ahpra’s news item of 22 August 2024 keeps practitioners under their code of conduct. Meridian Lawyers, in a summary dated 18 October 2024, reads the guidance as requiring the practitioner to apply human judgment to any AI output.
Is a chatbot on the website a medical device?
It depends on its intended purpose, not on the word “chatbot”. The TGA gives a chatbot that suggests, delivers or monitors treatment as one of its own examples of a medical device (page updated 5 February 2026).
A chat that gives approved practice facts, takes a message and hands the enquiry to the front desk does a different job. If an update later adds diagnosis or treatment suggestions, the TGA calls that scope creep: the tool has changed its intended purpose and needs ARTG inclusion before that version is supplied.
Shoutout Digital’s chat agent hands enquiries to the front desk and does not book. This is general information, not legal advice.
Can the receptionist use ChatGPT to draft a letter?
Yes for a letter with no patient details in it. No for anything that carries a patient’s name, date of birth or history into a personal or free account.
The OAIC (published 21 October 2024, updated 17 January 2025) says personal information put into an AI system is still personal information. Its best-practice line is not to enter personal information, particularly sensitive information, into a publicly available generative AI tool. Clinic Comply restated that for practices on 6 October 2026.
A person must still read the draft before it goes out. ADA Policy 6.34 keeps generative AI to low-risk drafting that a person checks.
Can AI write our Google Ads?
It can draft them. A person who understands the advertising rules must approve every public sentence.
Section 133 of the National Law still applies to an ad a model wrote. The ad must not be false or misleading, must not use a testimonial about a clinical aspect of care, must not create an unreasonable expectation of benefit, must not encourage unnecessary use, and must state the terms of any gift or discount. Ahpra’s guidelines page was reviewed on 2 September 2025.
The same rules bind an ad, a landing page and a reply to a review, whoever or whatever wrote the words.
Will AI replace the dentist or the front desk?
AI is a helper with a person who signs.
ADA Policy 6.34 says clinical AI is supervised by a dental practitioner and is not actioned by the patient alone. The ARTG entries describe tools that act as a second reader, chart a tooth or highlight a finding for the dentist. On the phone, an AI chat can answer approved facts and capture a message, then hand emergencies and treatment questions to a person.
So the dentist still diagnoses, the front desk still looks after the patient, and a person still approves the ad.
Do we have to tell patients we use AI?
Tell them. Ahpra lists transparency and informed consent among its principles for AI in healthcare (news item, 22 August 2024), so tell patients in plain words when their information goes into a tool.
The OAIC adds two rules for a public-facing chat: identify it clearly as a chatbot, and put a plain line about the AI in your privacy policy and collection notice (guidance updated 17 January 2025).
ADA Policy 6.34, clause 2.14, also says practitioners should tell patients about the risk of asking a public generative AI tool for a diagnosis or a treatment plan. For the exact wording in your practice, ask an independent lawyer.
Where should a practice start if it wants one tool, not five?
Start with the job that costs the most time, and match the tool to that job. Do not start with the tool.
For a second reader on radiographs, choose a tool whose ARTG entry covers that job, and accept that the dentist still signs the finding. For the ads account, the pages and the follow-up, a marketing agent fits, with a person approving public words and spend. For the phone, a chat that hands over to the front desk fits. A public chatbot is the wrong product for the patient record.
Whichever you pick, ask for the ARTG entry if the tool is a device, and ask who in your practice signs.
Sources, with dates
How this page was made. Drafted with AI assistance from the sources below, opened on the dates shown.
- Australian Dental Association: Policy 6.34, artificial intelligence in dentistryDocument version November 2023, amended 17 November 2023; opened 12 October 2026
- TGA: artificial intelligence and medical device software regulationPage last updated 5 February 2026
- TGA: AI-enabled medical devices in the ARTGList announced 4 September 2025; retrieved 12 October 2026
- TGA: digital scribesPage dated 30 January 2026
- OAIC: guidance on privacy and the use of commercially available AI productsPublished 21 October 2024, updated 17 January 2025
- Ahpra: AI in healthcare and the code of conductNews item, 22 August 2024
- Ahpra: guidelines for advertising a regulated health serviceIn force since 14 December 2020; page reviewed 2 September 2025
- Meridian Lawyers: Ahpra issues guidance on using artificial intelligence18 October 2024
- Australian Commission on Safety and Quality in Health Care: AI Clinical Use GuideVersion 1.0, August 2025
- Clinic Comply: AI and privacy compliance for healthcare practices6 October 2026
Related reading. See the difference between AI agents and chatbots, how AI agents handle patient data in a dental practice, and how AI live chat hands enquiries to your front desk.
Published . Last updated .