Ethical Considerations for Psychologists Using AI Clinical Documentation in Australia

The Regulatory Framework

APS Code of Ethics

The Australian Psychological Society's Code of Ethics (2007) does not specifically address AI tools — it predates their clinical application. However, several general principles apply directly:

General Principle A: Respect for the Rights and Dignity of People and Peoples

• Standard A.3 (Privacy): Psychologists must respect clients' right to privacy and collect only information relevant to the service being provided.
• Standard A.5 (Informed Consent): Clients must be informed about the nature, purpose, and foreseeable consequences of the service, including how their information will be handled.

General Principle B: Propriety

• Standard B.1 (Competence): Psychologists must practise within their areas of competence, which includes understanding the tools they use.
• Standard B.6 (Record keeping): Psychologists must maintain adequate records of professional services.

General Principle C: Integrity

• Standard C.4 (Accuracy): Psychologists must take reasonable steps to ensure the accuracy of information they record and communicate.

AHPRA and the Psychology Board of Australia

The Psychology Board of Australia's Code of Conduct (2025) and Professional Competencies for Psychologists (2025), both effective 1 December 2025, establish record-keeping requirements that are directly relevant:

• Records must be accurate, legible, and contemporaneous
• Records must be stored securely and confidentially
• Practitioners must be able to demonstrate the basis for clinical decisions
• Records must be maintained for 7 years after the last date of service (or until the client turns 25 if they were a minor)

The 2025 Professional Competencies include greater emphasis on digital competence, providing a clear framework for evaluating AI documentation tools in psychological practice.

State and Territory Legislation

Surveillance and recording laws vary by state and territory and are critically important for psychologists using ambient recording:

• NSW (Surveillance Devices Act 2007): All-party consent required for recording private conversations. The psychologist and client must both consent.
• Victoria (Surveillance Devices Act 1999): One-party consent for recording private conversations (the recording party must be a party to the conversation).
• Queensland (Invasion of Privacy Act 1971): One-party consent.
• Western Australia (Surveillance Devices Act 1998): All-party consent.
• South Australia (Surveillance Devices Act 2016): All-party consent.
• Tasmania (Listening Devices Act 1991): All-party consent.
• ACT (Listening Devices Act 1992): All-party consent.
• NT (Surveillance Devices Act 2007): All-party consent.

In most Australian jurisdictions, you cannot legally record a therapy session without the client's knowledge and consent. Even in Victoria and Queensland where one-party consent technically applies, ethical obligations under the APS Code require informed consent regardless of what the law permits.

Informed Consent for Recording

The most fundamental ethical requirement for psychologists using AI documentation is informed consent. This applies whether you are using ambient recording (capturing the full session) or dictation (recording your own voice post-session).

Ambient Recording Consent

If you record therapy sessions for documentation purposes, clients must be informed:

1. What is being recorded. The audio of the therapy session.
2. Why it is being recorded. For clinical documentation purposes — to generate session notes.
3. How the recording is processed. It is transcribed by a speech recognition system and then used by an AI language model to generate clinical notes.
4. What happens to the recording. Whether it is stored, for how long, and when it is deleted.
5. Who has access. Which third-party systems process the audio and where data is stored.
6. Their right to refuse. Clients must be able to decline recording without this affecting their access to treatment.

Dictation Consent

Dictation raises fewer consent issues because you are recording your own voice summarising the session, not recording the client. However, transparency requires that clients know you use AI tools to assist with documentation. A brief mention in your informed consent documentation is appropriate.

Sample Consent Wording

The following is a starting point — adapt it to your specific practice and the tools you use:

'I use an AI-assisted documentation tool to support my clinical note-taking. [If using ambient recording: With your consent, I record our sessions using a secure recording tool. The audio is processed by a speech recognition system to create a transcript, which is then used to generate clinical notes. The audio recording is deleted immediately after transcription and is not permanently stored. / If using dictation: After our sessions, I dictate a clinical summary which is processed by an AI system to generate structured notes.] The AI-generated notes are reviewed and edited by me to ensure accuracy before they become part of your clinical record. Your data is processed within Australian servers and handled in accordance with the Australian Privacy Principles. You may decline the use of this tool at any time without affecting your treatment.'

Responsibility to Review AI Output

This is perhaps the most critical ethical consideration. When an AI generates a clinical note, the psychologist who signs off on that note is professionally responsible for its accuracy and completeness.

The Automation Bias Risk

Automation bias is the tendency to over-trust automated systems and accept their output without adequate scrutiny. In the context of AI documentation, this manifests as:

• Accepting AI-generated notes without reading them carefully
• Assuming the AI has captured everything accurately because 'it usually does'
• Failing to notice when the AI has added content that was not part of the original session
• Missing instances where the AI has omitted clinically significant information

Research in aviation, radiology, and other fields consistently demonstrates that professionals are prone to automation bias, particularly when they are fatigued or time-pressured — exactly the conditions under which clinicians complete their documentation.

Practical Mitigation Strategies

1. Review every note before finalising. This sounds obvious, but the time pressure that drove you to adopt AI documentation in the first place creates pressure to skip the review step. Build review time into your workflow — even 2 to 3 minutes per note.

2. Compare the note to your clinical memory. Does the note accurately reflect what happened in the session? Are the presenting concerns, interventions, and outcomes correct? Is the tone appropriate?

3. Check for AI fabrication (confabulation). Large language models can occasionally generate plausible-sounding clinical content that was not part of the original session. This is particularly problematic in clinical documentation because fabricated content could mislead future treating practitioners. Look for specifics that seem unfamiliar — dates, scores, intervention names, or clinical observations you do not recall making.

4. Verify clinical accuracy. If the AI generated a note that mentions a diagnosis, ensure the diagnosis is correct. If it mentions a treatment modality, ensure you actually used that modality. If it mentions outcome scores, verify the numbers.

5. Watch for inappropriate inferences. AI models may infer information that was implied but not stated. For example, if a client mentioned feeling 'terrible,' the AI might document 'client reports depressed mood' — which may or may not be an accurate clinical characterisation.

Maintaining Clinical Record Accuracy

The Psychology Board requires that records be accurate. When using AI-generated documentation, several specific accuracy concerns arise:

Diagnostic Precision

AI-generated notes may use diagnostic language loosely. If a client describes symptoms of low mood, the AI might generate 'client presents with depressive symptoms consistent with Major Depressive Disorder' when in fact no formal diagnostic assessment has been conducted. Always check that diagnostic statements in AI-generated notes reflect your actual clinical assessment.

Treatment Modality Accuracy

If you described an intervention in general terms, the AI may assign it a specific treatment modality label. For example, a behavioural activation intervention might be labelled as 'CBT' or a mindfulness exercise might be attributed to 'ACT.' Ensure the treatment modality descriptors in the note match what you actually practise.

Temporal Accuracy

AI-generated notes may occasionally confuse the temporal sequence of events within a session or introduce information from previous sessions (if context from previous notes was provided). Check that the chronology is correct.

Telehealth Considerations

Telehealth sessions introduce additional complexity for AI documentation:

Recording telehealth sessions. Recording a video call may fall under different (or additional) legislation compared to in-person recordings. Some telehealth platforms have their own recording policies that may conflict with third-party AI tools.

Network security. Audio from telehealth sessions transmitted to an AI platform travels through additional network infrastructure. Ensure your AI tool uses encryption in transit and at rest.

Client location. If your client is in a different state or territory, the recording laws of their jurisdiction may apply. A client in NSW (all-party consent) consulting with a psychologist in Victoria (one-party consent) creates jurisdictional complexity. The conservative approach is to always obtain explicit consent.

Client Access to AI-Generated Records

Under the Australian Privacy Principles (APP 12), individuals have the right to access their personal information held by an organisation. This means clients can request access to their clinical records, including AI-generated notes.

This creates a practical consideration: AI-generated notes should be written as if the client might read them. Avoid pejorative language, unsupported interpretations, or notes that could be harmful to the therapeutic relationship if the client saw them.

This is good clinical practice regardless of AI involvement, but the speed at which AI generates notes can sometimes produce language that a psychologist, writing manually, would have self-edited.

Legal Liability

If an AI-generated note contains an error that leads to adverse client outcomes — for example, an incorrect medication is documented, leading a GP to prescribe inappropriately — the psychologist who approved the note bears professional responsibility.

The AI platform is a tool, not a practitioner. It does not hold professional registration, professional indemnity insurance, or a duty of care to the client. The duty of care rests entirely with the psychologist.

This is not fundamentally different from the liability that exists when using any documentation tool (voice-to-text software, templates, or even a secretary who types your notes). The practitioner is always responsible for the final record.

However, the sophistication of AI output may create a false sense of security. A note that reads fluently and uses appropriate clinical language may contain errors that are harder to detect than obvious typos or formatting issues.

Practical Recommendations

1. Include AI documentation in your informed consent. Update your consent forms to disclose the use of AI tools, the recording process, data handling, and the client's right to opt out.

2. Build review time into your workflow. Do not treat AI-generated notes as final. Every note requires human review by the treating psychologist before it becomes part of the clinical record.

3. Choose a platform with transparent data practices. Know where your data is processed, whether audio is retained, and how the platform handles data breaches. Platforms like Grounded Scribe process data within Australian infrastructure and delete audio immediately after transcription.

4. Maintain your clinical skills. AI documentation should supplement your note-writing abilities, not replace them. If you lose the ability to write clinical notes without AI assistance, you have become overly dependent on the tool.

5. Stay informed about regulatory developments. The Psychology Board of Australia, APS, and AHPRA are likely to issue specific guidance on AI documentation tools as adoption increases. Monitor their communications and adapt your practice accordingly.

6. Document your AI review process. Consider adding a brief notation to your records indicating that AI-generated notes have been reviewed and approved — for example, a standard footer or a note in your practice management system.

7. Discuss AI use in supervision. If you are a supervised psychologist or a supervisor, include AI documentation practices in supervision discussions. This is a rapidly evolving area where professional guidance is valuable.

Looking Ahead

AI clinical documentation is not a passing trend. The tools will continue to improve in accuracy, and their adoption across Australian psychology practice will continue to grow. The ethical framework for their use is still developing, but the existing principles — informed consent, accuracy, professional responsibility, client welfare, and transparency — provide a robust foundation.

The psychologists who navigate this transition most successfully will be those who embrace the efficiency gains while maintaining rigorous ethical standards and professional judgement. The tool serves the practitioner; the practitioner serves the client.

Important Disclaimer

*This article is for informational purposes only and does not constitute legal, clinical, or regulatory advice. Grounded Scribe is a documentation tool — it does not provide legal guidance or ensure compliance with any specific legislative, regulatory, or registration body requirements. Practitioners are solely responsible for ensuring their documentation meets the standards of their registration board, employer, and applicable legislation. All AI-generated content must be reviewed, edited, and approved by the practitioner before it becomes part of the clinical record. For medico-legal, child protection, or tribunal documentation, always seek independent legal and professional advice relevant to your jurisdiction and specific circumstances.*

---