AI & Mental Health - How, Why and Beyond

Do I use it in clinical practice?

Yes, I have adopted the use of NovoNote during sessions to assist me with taking notes for sessions.

NovoNote is a platform within Novopsych.

About NovoNote: NovoNote is produced by NovoPsych, an Australian company founded in 2012. NovoNote is a secure and encrypted platform that converts the audio from our consultations into clinical summaries.

Compliance and security: NovoNote complies with the Australian Privacy Principles and is compliant with AHPRA and HIPAA standards. It adheres to industry encryption and security protocols. Importantly, audio recordings from sessions are not saved; instead, they are transformed into a transcript of the session and then discarded. NovoNote stores data on a secure server in Australia and their data processors do not use your data to train AI models.

For information about Novonote's security, follow this link.

What is saved: As healthcare providers, we keep notes of our contact with you to aid your treatment, and to comply with our ethical and legal obligations. The AI note taker NovoNote, helps by transcribing the session and using that transcription to provide a written summary of the session, which is saved in your confidential client file. The transcript of the session is then deleted within 3 months, and does not make up part of your patient file. Audio recordings of the session are not saved.

Why do I use it?

For context, I may typically spend an additional 20 minutes on writing up the initial 75 minute initial appointment for a client, and anywhere between 5-15 minutes for a 50-minute ongoing appointment. I typically see 5 clients a day, which means on average 10 mins x 5 clients = 50 minutes of unpaid time to complete my paperwork. That is not inclusive of reports to GP for Medicare, support letters, time to obtain collateral information from other health practitioners involved in your care, and other administrative tasks (e.g., following up on medicare rebate processing, payments that did not process, rescheduling appointments).

So, the biggest reason I have adopted the use of AI note taker is to reduce my administrative time. Good documentation is crucial for safe care for my clients, as well as meeting my legal and ethical requirements as a psychologist. As a psychologist, we live with the looming possibility that our notes may one day be subpoenaed by the court and being cross-examined for an overlooked typo (yes, that is a bit of catastrophising!). All that to say, it takes a lot of time because it deserves that effort. However, most psychologists are more interested in doing the real work - actually engaging with our clients, engaging in resource building - looking up the latest research or reading a book chapter relevant to our current clients, or spending time nourishing ourselves so we can actually continue to do good therapy.

The second reason is that it's really hard to stay present to you in session if I am worried about forgetting to note down certain details of what you've shared (the name of that horrible manager!) and clinical observations/hypotheses I am making as I am listening to you speak.

Your right as a client

No one will be pressured to opt-in. If you are uncomfortable with the use of AI-note taking in session, you are absolutely within your right to opt-out. Your consent is usually sought verbally during the first session, however, please feel free to raise this at any time. You are also encouraged to withdraw your consent at any time should you change your mind.

Therapy and AI, Therapy from AI?

The truth is that AI chatbots and digital companions have emerged suddenly and quickly. This means that the research on the impact on people's mental health is emerging but lacking.

There has been preliminary evidence around emotional relief with young people (Li J, et al 2025) - noting that these chatbots were varied and some appear to be within mental health specific platforms and not AI chats like Claude or ChatGPT. On the other hand, Carlbring & Andersson 2025, highlights significant failures of AI based on LLMs (large language model) - basically ChatGPT, Claude etc - to (1) collude with delusional beliefs (anecdotally, there have been report of psychosis caused by AI), and (2) enable suicidal ideation by providing information how to plan suicide, and (3) overly agreeable with users to be maintain user satisfaction, and (4) is stigmatizing toward people with certain diagnoses.

Outside of mental health specific literature, LLM's racism has also been well documented.

Jose et al 2025 writes about cognitive offloading, which refers to using an external aid (in this case, AI) to complete takes requiring thinking. It further reduces opportunity to train our memory and problem-solving skills.

LLMs directly and negatively impacts on the following things that I believe to be crucial in maintaining psychological well-being:

• Critical thinking

• Patience (and the tolerance to delay gratification)

• Tolerating frustration and boredom e.g., learning to tolerate the discomfort of not knowing something yet, of trying to figure something new slowly

• Learning through trial and error e.g., drafting that text to your friend about something hard to bring up rather than off-loading to an LLM to write you a script and do it 'right' the first time.

• Metacognition: our ability to identify our own thoughts, engage in deliberate self-reflection, create meaning about events in our life, create hypotheses about the cause and effect about those events, and adjust our logic based on new information.

I have to admit that I was drawn to it when it first was made available. In particular, I saw a way for clients to maybe have something (that was free and accessible) in between sessions - something that could help. However, over time, the impacts I have witnessed in my therapy room, in conversations in personal spaces and online, and not to mention the impact on the environment! My views have changed dramatically.

In the last few months, I have made a personal commitment to limit the use of AI/LLMs as much as possible (with my one exception of NovoNote as discussed above).

I would never judge any clients for using AI. However, I may choose to encourage you to reflect on its impacts on you where it relates to the reason you are engaged with me in therapy.

This post is written with the intention to support current/future clients to understand how AI will/won't be used in therapy with me, to share some current information about the use of AI and how it relates to psychological well-being, and also to share something personal around my views on AI for anyone who wants to know.

For the full disclaimer on blog content, please read this post.