Artificial intelligence is reshaping how psychologists work

Psychologists are increasingly turning to tools powered by artificial intelligence (AI) to streamline their practice—about 1 in 10 use it at least monthly for note-taking and other administrative work, according to the 2024 Practitioner Pulse Survey by APA and APA Services. But many remain skeptical, with 71% reporting they’ve never used AI in their practice.

“Historically, technology is something that psychologists have been rightly wary of,” said Elizabeth Stuart, PhD, a pediatric neuropsychologist at Miller Children’s Hospital in California who uses AI for notetaking, clinical insights, and practice management. This caution often stems from a deep commitment to protecting patient privacy and adhering to ethical standards. “But AI is here. As a field, we need to join in, or we’re going to get left behind.”

AI isn’t just another tech trend—it’s reshaping how psychologists do their work. From streamlining paperwork to highlighting trends in patient data, it offers powerful ways to elevate your practice. In June 2025, APA released Ethical Guidance for AI in the Professional Practice of Health Service Psychology, which outlines key considerations for health service psychologists who use AI in professional practice. The first-of-its-kind report draws on the APA Ethics Code and other standards and guidelines to advise psychologists on how to leverage AI’s potential ethically and responsibly. Here’s further advice from experts on how to tap into that potential while staying aligned with the ethics of clinical care.

Multiple levels of support

Many AI tools for psychology practitioners were originally designed to tackle one major pain point: clinical documentation. Now, many have evolved into full-service platforms that offer scheduling, billing, claims management, and more.

“No one gets into mental health care to do paperwork, but administration is a massive part of the job and something that can drain a lot of energy,” said Damien Adler, a psychologist based in Victoria, Australia, who cofounded the global practice management software company Zanda Healthopens in new window after seeing how technology could boost efficiency in his own practice. “Fortunately, AI can now significantly reduce that burden.”

Zanda’s AI companion offers multiple levels of support, tailored to how much AI integration a therapy practice prefers. “Refine” is one layer of support—and a good choice for practitioners with concerns about privacy. Rather than listening to a session, it converts a therapist’s own rough notes into reports that comply with payer or regulatory requirements. The “Transcribe” function listens in and generates a clinical note that can later be edited or customized. “Insights” reviews patient records, attendance patterns, and documented goals to provide clinicians with tailored clinical observations and treatment suggestions.

The progress notes tool Uphealopens in new window caters specifically to mental health professionals, so it’s tuned to pick up themes that mean more in therapy—such as personal growth and coping—than they might in other types of appointments that require transcription, such as speech therapy sessions. Reimbursement guidelines around medical necessity from many major insurance payers are infused into the model. These guidelines inform what’s added to various sections of a report, note, or document to help ensure alignment with payer requirements.

“The threat of a claw-back [a demand to repay insurance funds] on a small practice or individual provider is incredibly daunting, but there can be a lot of ambiguity about precisely what documentation payers expect to see,” said Ted Faneuff, LCSW, Upheal’s head of clinical operations.

After launching an in-app demo client called Alex—designed to let therapists experiment with note output without using real patient data—many therapists shared enthusiasm for using AI clients in clinical training, seeing it as a low-stakes way to practice therapy techniques. While Alex is currently not programmed to do this, the excitement is there, Faneuff said.

Platforms built for health care professionals should always provide transparency on their data usage and training policies, Stuart said. Upheal, for example, never sells patient data or uses it to train its model. BastionGPTopens in new window, a generative AI tool similar to ChatGPT that assists with note-taking and other tasks, is trained using data from credible sources, such as WebMD. The tool is also tuned to reduce hallucinations—AI-generated responses that sound plausible but are incorrect.

“In sensitive fields like mental health, accuracy matters just as much as ease of use,” said Josh Spencer, a cybersecurity expert and founder of BastionGPT. “AI needs to support psychologists by not just reducing documentation burden, but by actually improving the quality of their documentation.”

One of the most exciting applications of AI is to help clinicians tailor treatment in real time. By analyzing session themes, tone of voice, attendance patterns, wearable data, and more, therapeutic algorithms can increasingly flag signs of relapse risk, treatment dropout, or even potential suicide attempts. While AI can alert a clinician to a pattern, it’s ultimately the provider who is responsible for interpreting those signals, adding context, and deciding the next steps in patient care.

In addition to flagging risks, AI tools can also help clinicians fine-tune treatment. For example, a system might recommend switching to narrative therapy when a patient isn’t progressing with more structured approaches.

The following AI tools for notetaking, clinical support, and practice management offer a good starting point, according to Stuart and other practitioners:

• Mentalycopens in new window : Note generator built specifically for behavioral health, which uses an encrypted connection and never stores recordings
• Blueprintopens in new window : Focuses on offering clinical insights and measurement-based care by automating progress notes, drafting treatment plans, and surfacing actionable insights before, during, and after sessions
• Assessment Assistantopens in new window : Supports assessment providers with reporting, record review, and other evaluation, using a language model informed only by clinical sources

Using AI in your practice can save time in the long run—but psychologists say that getting it right requires a thoughtful setup, from choosing the best tool to safeguarding patient privacy.

“A measured approach to adoption that considers both compliance and the perspectives of patients and staff is essential for AI success,” said Mollie Cummins, PhD, RN, a professor in the College of Nursing at the University of Utah, who blogs about AI best practices at telehealth.orgopens in new window. She and other experts suggest that clinicians:

Select the right tool. Choose a HIPAA-compliant tool that prioritizes data security, including by encrypting data and logging user activity, Cummins advises ( see her “Mastering AI Documentation in Telehealth” blog post for more tipsopens in new window ). Other factors to consider include how customizable a tool is and whether it was trained using terminology and documents specific to behavioral health.

APA offers a step-by-step guide and companion checklist to help practitioners decide whether a given AI tool is right for your practice.

Practice prompt engineering. Crafting clear, well-structured prompts can help you unlock significantly more value from AI. Charmain Jackman, PhD, founder of InnoPsychopens in new window and a member of APA’s Mental Health Technology Advisory Committee, explains how giving AI software a role, specific instructions, and context can yield better results . Jackman uses AI to tailor marketing emails, brainstorm social media content, and break down complex psychological topics for a general audience.

Stuart suggests a few additional ways to refine outputs: provide further instructions (e.g., “ignore information about the patient’s pet”), clarify format (e.g., “write this in Subjective, Objective, Assessment, Plan (SOAP) note format”), and give guidance on tone (e.g., “use short sentences and language that could be understood by an 18-year-old”).

Be proactive about risk. Many providers shy away from AI because they are unsure how to protect patient privacy and security while using it. Cummins recommends a few key actions ( see her “AI Scribing in Telehealth” blog post for more adviceopens in new window ):

• Make sure patient data is stored and transmitted securely. Use a VPN, HIPAA-compliant cloud storage, and cybersecurity tools that monitor threats in real time. If you work at a larger institution, contact your security team for support.
• Verify that AI-generated responses are accurate. Providers are ultimately responsible for any documentation they sign or decision they make, so your workflow should include a thorough review of any AI outputs.
• Make sure your staff understands and uses cybersecurity best practices, including strong passwords, a secure Wi-Fi network, and multifactor authentication. In case of a security breach, create a plan for your practice (find out more about how to prevent or recover from a cyberattack).
• Before adopting a tool, take time to understand how it handles patient data and how the underlying model was trained.

Always get verbal or written consent from patients before using AI tools during a session. Explain a tool’s risks and benefits in plain language—and be prepared to proceed without the tool if a patient opts out, experts advise. “Compared to [psychology], Silicon Valley moves very fast,” Stuart said. “It’s imperative that we integrate AI tools into our practice, but let’s be careful about how and when we use them.”