AI and dental patient communication under GDC standards

A dental practice looking for a GDC rule on AI answering the phone will not find one. What it will find is a set of standards written for people, which apply with full force to software doing the same job, and one recent piece of GDC research that shows how early the profession still is. The standards that decide this are about introducing yourself, communicating in a way patients understand, and keeping records that are complete, confidential and used only for their original purpose.

No standard specific to AI exists. The GDC published a rapid evidence assessment on artificial intelligence in dental service provision on 12 August 2025, carried out by Peninsula Dental School at the University of Plymouth across 45 international studies from 2020 onwards. That review covers clinical uses: robotics in implant surgery, deep learning for caries detection, and machine learning predicting childhood caries. It found no UK studies among them, and named gaps in best practice guidelines, UK specific applications, ethical considerations and data protection. Nothing in it addresses a system that speaks to patients. This is general information about published GDC material and not clinical or dental advice.

The GDC review of research into AI and dental service provision, published 12 August 2025, is a useful marker of where the evidence stands. Peninsula Dental School at the University of Plymouth examined 45 international studies from 2020 onwards across twelve countries, with China contributing the most and no UK studies included at all. The applications it found cluster in implant robotics, deep learning for caries detection and machine learning that predicts childhood caries. The GDC records the review's own conclusion that significant gaps remain in best practice guidelines, large-scale trials, UK-specific applications, ethical considerations, data protection, and the effect on equality, diversity and inclusion.

Read that as the regulator publishing a map with the front of the practice left blank. It is a good reason for caution and a poor reason for inaction, because the standards that govern patient communication were already in place and never depended on the technology.

The general communication and records standards, written for people and binding on whatever a practice uses to do the same job. Standard 2.1 requires effective communication that takes patients' individual views and communication needs into account. Guidance 2.3.1 asks a professional to introduce themselves and explain their role so the patient knows how they are involved in their care. Principle Four covers records and confidentiality. The patient expectations listed under Principle Two include knowing the names of those providing their care.

The GDC Standards for the Dental Team open Principle Two with what patients expect, and the list includes receiving information they can understand and knowing the names of those providing their care. Standard 2.1 requires effective communication that takes patients' individual views and communication needs into account, and guidance 2.1.1 asks that patients are treated as individuals with their communication needs and preferences respected where possible. Guidance 2.3.3 goes further into practical accessibility, naming plain language over jargon, interpreters, and support for patients who use sign language or hearing aids.

Every one of those is a design requirement for a system that answers calls. A caller who cannot make themselves understood by an automated line, and has no way through to a person, is a standard 2.1 problem long before it is a technology problem.

The standards point firmly that way without naming AI. Principle Two of the GDC standards lists, among the things patients expect, to know the names of those providing their care, and guidance 2.3.1 asks professionals to introduce themselves and explain their role. A patient who believes a member of the practice team is handling their call, when the caller is speaking to software, has been given a false picture of their own care. A single sentence at the start of the call settles it.

This is the point where the absence of an AI-specific rule stops mattering. Guidance 2.3.1 asks a professional to introduce themselves and explain their role so the patient knows how they are involved in their care, and the expectation that patients know who is providing their care sits at the top of the principle. A system that presents as a receptionist without saying what it is leaves the patient with a false understanding of their own care, which the standards address plainly even though they were written before this technology arrived.

Only with the patient's permission. GDC guidance 4.2.9 extends the duty of confidentiality to recordings of patients, audio recordings among them, and states that you must not make any recordings or images without the patient's permission. That condition is settled before any question about the technology. Once a recording exists, guidance 4.1.1 places audio or visual recordings of consultations in a patient's records where they are available, so it is retained and secured as part of the record. Standard 4.2 requires that patient information stays confidential and is used only for the purpose for which it was given, which is the standard a supplier's model training terms have to be measured against. Guidance 4.1.3 requires professionals to meet their responsibilities for patient information under current legislation, which brings UK GDPR with it.

The retention consequence catches practices out. Principle Four guidance 4.1.1 lists audio or visual recordings of consultations among the things that form part of a patient's records where they are available, alongside radiographs, consent forms and referral letters. A call recording that captures clinical discussion is therefore a record to be retained and secured, and standard 4.2 requires that patient information stays confidential and is used only for the purpose for which it was given. A supplier clause permitting model training on call content is the direct collision with that standard, and it is the first clause to read.

Guidance 4.2.9 is the item that decides whether a recording may be made at all, and it sits in the same principle. It extends the duty of confidentiality to recordings or images of patients, naming photographs, videos and audio recordings, originals and copies, and those made on a mobile phone, and it states that you must not make any recordings or images without the patient's permission. Guidance 4.2.2 adds that non-registered members of the dental team must understand the importance of confidentiality and keep patient information confidential at all times, which is the closest analogue the standards offer for a tool the practice has introduced into the same conversations.

The practice and the registered professionals in it. A Product News item in the British Dental Journal on 23 May 2025 stated that clinicians remain legally and ethically responsible for record keeping accuracy and must ensure any AI generated content is customised and critically reviewed, and warned that a discrepancy between a transcript and a recording could form the basis of a legal challenge. The GDC standards carry the same logic: they bind the registrant, and the choice of tool leaves the duty exactly where it was.

The British Dental Journal has covered this squarely. A Product News item on 23 May 2025 (Br Dent J 238, 826) surveys where AI is being used in dentistry, placing appointment scheduling and data handling under practice management, and observing that its largest current use is in record-keeping and patient communication. Its warning is the part to keep: clinicians remain legally and ethically responsible for record-keeping accuracy and must ensure AI-generated content is customised and critically reviewed. It also flags a risk specific to automated call handling, that a discrepancy between a transcript and a recording could form the basis of a legal challenge.

Worth being precise about that source: it is a Product News item rather than peer-reviewed research, and it reflects a medico-legal defence organisation's view. It is cited here because it is a real, dated, checkable statement of professional accountability from the profession's own journal, and because we could verify no British Dental Journal research specifically examining conversational AI on a dental practice telephone. If that study exists, it did not surface.

Say what the system is at the start of the call. Keep a route to a person for anyone who needs one. Treat recordings as part of the record, with the retention and security that implies. Read the supplier's terms against standard 4.2 and refuse training on patient content. Keep clinical questions away from the system entirely, because nothing in the GDC's own evidence review supports putting them there.

That is the scope an AI employee is built to in a dental practice: booking, rescheduling, recalls and reminders in wording the practice approved, with anything clinical going to a named person. The same boundary across four markets is in the rules on regulated AI intake, the reminder evidence sits in reminders that actually get answered, and what a practice would recover is on the ROI calculator.