AI in Psychological Practice: What a New Survey Found About Use, Trust, Ethics, and Training
Key Points (For Readers in a Rush)
In a survey in collaboration with the National Register of Health Psychologists AI and Emerging Technologies Committee, 56.9% of licensed health service psychologists reported using AI for work in the past six months.
Among AI users, 51.4% reported using AI weekly or daily.
Psychologists were more likely to use AI for work-related questions, plain-language explanations, presentation content, and progress notes than for test interpretation or report writing.
Only 20.0% of respondents agreed that they trust AI, but 54.8% agreed that AI is helpful in practice.
73.0% believed practitioners should routinely disclose AI use to clients, yet the median reported disclosure rate among AI users was only 10%.
81.6% of respondents expressed interest in AI-related training or resources from the National Register of Health Service Psychologists.
Introduction
AI in psychological practice is no longer a future issue. It is already part of how many psychologists work.
This post is a more in-depth follow-up to my earlier blog post, AI Use Among Health Service Psychologists: Current Use and Changes from 2024 to 2025. That earlier post summarized broad trends in AI use across psychology. This post goes deeper into the descriptive survey findings now that the preprint is available and the manuscript is under review.
The preprint is Helpful but Not Trusted: Artificial Intelligence Adoption and Ethical Concerns Among Health Service Psychologists.
In the study, my colleagues and I surveyed 518 members of the National Register of Health Service Psychologists about their AI use, ethical concerns, privacy practices, disclosure behaviors, barriers to adoption, and training needs. The findings offer a descriptive snapshot of where the profession appears to be right now: many psychologists are using AI, many find it helpful, but trust remains low and ethical concerns remain substantial.
That mix is important. The data do not suggest a profession blindly rushing into AI. They also do not suggest a profession rejecting AI outright. Instead, psychologists appear to be experimenting, using AI selectively, and asking for clearer guidance.
How Many Psychologists Are Using AI?
The first major finding is straightforward: AI use among psychologists is already common.
Among respondents with valid AI-use data, 56.9% reported using AI for work purposes in the past six months. By comparison, 43.1% reported no work-related AI use during that same period. Personal use was even higher: 71.2% reported some personal AI use in the past six months.
Among those using AI professionally, use was often regular:
19.0% used AI less than once per month.
9.9% used AI about once per month.
8.8% used AI about twice per month.
10.9% used AI about three times per month.
33.0% used AI weekly.
18.4% used AI daily.
Taken together, 51.4% of AI users reported weekly or daily work-related use. That matters because occasional experimentation is different from routine workflow integration.
Psychologists also reported saving time. Among AI users, the average reported time savings was 3.89 hours per week, with a median of 2 hours per week. That is likely one reason AI use is continuing even though many psychologists remain cautious.
What Are Psychologists Using AI For?
The most common AI uses were practical, informational, and communication-focused.
Among AI users:
39.5% used AI to answer work-related questions.
35.0% used it to generate easy-to-understand explanations.
33.3% used it to create presentation content.
31.6% used it to write session or progress notes.
25.2% used it to generate psychoeducational materials.
24.5% used it to generate emails.
18.7% used it to generate summaries.
These findings suggest that many psychologists are using AI in ways that support communication, efficiency, and administrative work.
More clinically sensitive uses were less common:
21.1% used AI to generate clinical recommendations.
16.7% used it to generate treatment plans.
14.6% used it to generate treatment goals.
14.6% used it to write reports.
13.9% used it to transcribe sessions.
7.5% used it to interpret test results.
6.8% used it to translate materials.
4.4% used it for billing or coding.
This pattern is important. Psychologists seem to be drawing informal boundaries. They are more willing to use AI for tasks that support their work and less willing to use it for tasks that move closer to clinical judgment, assessment interpretation, or formal professional documentation.
Psychologists Are Using AI, But Many Do Not Trust It
One of the most interesting findings was the gap between usefulness and trust.
Only 20.0% of respondents agreed with the statement, “I trust AI.” By contrast, 62.9% disagreed, and 17.1% were neutral.
At the same time, 54.8% agreed that AI is helpful in practice. Only 22.9% disagreed, and 22.3% were neutral.
That means many psychologists appear to see AI as useful without necessarily seeing it as trustworthy.
This is one of the central findings of the study. I think of it as a trust–utility gap: psychologists may use AI because it helps with certain tasks, but they remain skeptical about its accuracy, judgment, and appropriateness.
That skepticism may actually be protective. In psychological practice, low trust can encourage review, caution, verification, and professional oversight. The problem is not that psychologists distrust AI. The problem would be using AI without understanding where that distrust is warranted.
Psychologists Do Not Think AI Is Better Than Practitioners
The survey also asked psychologists whether AI performs better than practitioners in several areas.
Most respondents did not view AI as superior to psychologists:
26.1% agreed that AI is better than practitioners at some clinical tasks.
29.8% agreed that AI is better at writing reports.
27.7% agreed that AI is better at integrating service-related data.
24.7% agreed that AI is better at selecting recommendations.
In each case, agreement was limited. For example, 47.1% disagreed that AI is better at selecting recommendations, and 49.2% disagreed that AI is better than practitioners at some clinical tasks.
This is another important descriptive finding. Psychologists may find AI helpful, but they do not appear ready to hand over professional judgment.
Ethical Concerns Depend on the Task
The study also found that psychologists’ ethical concerns varied depending on the AI use case.
Respondents were more accepting of some uses than others:
49.0% agreed that it is ethical to use AI for treatment or intervention planning.
45.6% agreed that it is ethical to use AI to write treatment or intervention goals.
38.1% agreed that it is ethical to use AI to write session notes.
30.2% agreed that it is ethical to use AI to write reports.
29.2% agreed that it is ethical to use AI to interpret test scores and client data.
The strongest concerns were about assessment interpretation and report writing:
53.3% disagreed that it is ethical to use AI to interpret test scores and client data.
52.3% disagreed that it is ethical to use AI to write reports.
44.2% disagreed that it is ethical to use AI to write session notes.
This is one of the most useful findings for practice. The question is not simply, “Is AI ethical?” The better question is, “Which AI use, for which task, under which conditions, with what safeguards?”
A psychologist using AI to draft a plain-language handout is not doing the same thing as a psychologist using AI to interpret assessment data. These activities carry different levels of risk.
Training Is One of the Clearest Areas of Agreement
Although respondents had mixed views about specific AI uses, they strongly supported training.
76.5% agreed that practitioners should be trained to use AI in practice. Only 10.1% disagreed, while 13.4% were neutral.
That is a striking finding. Psychologists may disagree about which AI uses are appropriate, but most appear to agree that the profession needs training.
The study also asked where respondents were already getting AI-related training or information:
49.1% had used webinars or online training.
46.0% had discussed AI with colleagues or coworkers.
36.7% had consulted journal articles.
34.1% had used websites.
14.4% had used social media.
14.0% had attended in-person training.
10.8% had used podcasts.
10.2% had used videos.
15.2% reported no training or information-seeking related to AI ethics and responsible use.
This pattern suggests that many psychologists are learning about AI informally. Webinars, colleagues, websites, and self-directed reading are playing a major role. That is understandable, but it also creates risk. If colleagues themselves have not received formal training, informal guidance may spread uncertainty or inconsistent practices.
The demand for formal professional development was clear: 81.6% of respondents said they were interested in AI-related training or resources from the National Register of Health Service Psychologists.
Disclosure Is a Major Practice Gap
Disclosure may be one of the most important findings in the study.
A large majority of respondents believed AI use should be disclosed:
73.0% believed practitioners should routinely disclose AI use to clients.
19.1% endorsed conditional disclosure depending on the context or application.
That means more than 9 in 10 respondents believed AI disclosure is either routinely needed or needed in at least some circumstances.
Actual disclosure behavior looked very different.
Among AI users, the average reported disclosure rate was 40.67%, but the median was only 10%, and the most common response was 0%.
That combination is important. The mean is pulled upward by some practitioners who disclose frequently. But the median and mode suggest that many AI users disclose rarely or not at all.
This does not necessarily mean psychologists are intentionally hiding AI use. It may mean the profession lacks clear procedures.
Psychologists may be unsure whether disclosure is needed when AI helps draft an email, organize a treatment plan, summarize notes, generate psychoeducational materials, or support report writing. Without specific guidance, disclosure becomes inconsistent.
Privacy and HIPAA Compliance Remain Unclear
Privacy findings also deserve careful interpretation.
Among AI users:
84.4% reported that they had not entered protected health information into an AI tool.
13.3% reported that they had entered protected health information.
2.4% were unsure whether they had done so.
The fact that most AI users reported avoiding protected health information is encouraging. It suggests that many psychologists understand that standard AI tools may not be appropriate for identifiable client information.
At the same time, formal compliance mechanisms appeared uncommon:
9.2% of AI users reported using a HIPAA-compliant AI vendor.
7.1% reported that they or their organization had a business associate agreement with the vendor.
That is the part that needs attention.
The finding is not simply, “Psychologists are misusing AI.” A better interpretation is that psychologists are trying to navigate privacy concerns in an environment where tool-specific guidance, compliance pathways, and institutional policies are still developing.
Psychologists need practical answers to questions such as:
What counts as protected health information in an AI prompt?
What does de-identification actually require?
When is a HIPAA-compliant vendor needed?
When is a business associate agreement required?
How should AI use be documented?
What should clients be told?
These are not abstract questions. They affect everyday practice.
Why Some Psychologists Are Not Using AI
The study also looked at barriers among respondents who had not used AI professionally.
Their reasons were not mainly about cost or resistance to change. They were mostly ethical, legal, and clinical concerns.
Among non-users:
71.3% cited ethical concerns.
70.0% cited privacy or confidentiality concerns.
69.1% cited accuracy or reliability concerns.
53.4% cited legal uncertainty.
50.2% cited concern about AI replacing human judgment.
47.1% cited bias concerns.
40.4% cited unfamiliarity with AI tools.
39.0% cited lack of evidence-based research.
36.3% cited lack of training.
32.7% cited workflow integration challenges.
29.1% cited impact on client relationships.
23.3% cited lack of practice-specific tools.
22.9% cited resistance to changing established practice methods.
13.9% cited uncertainty about how to explain AI use to clients, patients, or parents.
7.6% cited cost or budget constraints.
6.3% cited employer restrictions or uncertainty.
This matters because non-use should not be dismissed as simple reluctance.
Many psychologists are not avoiding AI because they dislike technology. They are avoiding it because they are concerned about ethics, privacy, accuracy, legal ambiguity, bias, and professional judgment.
Those concerns are not barriers to responsible practice. They are signs that psychologists understand the stakes.
Practical Examples for Psychologists and Practice Leaders
The findings point to a need for task-specific decision-making.
Lower-Risk Example: Drafting Educational Materials
A psychologist might use AI to draft a plain-language handout about coping skills, sleep hygiene, or anxiety. This may be a reasonable use if the psychologist reviews the output for accuracy, tone, cultural responsiveness, and fit with the client.
This type of use aligns with the more common uses reported in the study, such as generating easy-to-understand explanations and psychoeducational materials.
Moderate-Risk Example: Drafting Session Notes
Session notes were a common AI use, with 31.6% of AI users reporting this application. But documentation raises privacy, consent, and compliance questions.
Before using AI for notes, psychologists should know whether the tool is HIPAA compliant, whether a business associate agreement is in place, how data are stored, and how the final note is reviewed.
Higher-Risk Example: Test Interpretation
Only 7.5% of AI users reported using AI to interpret test results. This lower rate fits with the ethical concern respondents reported: 53.3% disagreed that it is ethical to use AI to interpret test scores and client data.
This is an area where caution is especially appropriate. Test interpretation requires clinical training, contextual knowledge, cultural and linguistic awareness, and careful integration of multiple data sources.
Practice Leadership Example: Creating an AI Use Policy
A clinic or agency could use findings from this study to create a tiered AI policy.
For example:
Allowed with review: drafting general educational materials or presentation content.
Allowed only with approved tools and safeguards: clinical documentation or client-specific summaries.
Restricted or prohibited without stronger evidence and oversight: diagnostic impressions, assessment interpretation, or unsupervised report writing.
This kind of policy would be more useful than a vague statement that staff should “use AI responsibly.”
Ethical Considerations
AI use in psychological practice raises several ethical issues.
Privacy and Confidentiality
Psychologists need to know whether client information is being entered into an AI system, where that information goes, and whether the tool is appropriate for clinical use.
Disclosure and Consent
The study found strong support for disclosure, but inconsistent reported practice. This suggests a need for clearer informed consent language and more specific professional guidance.
Bias and Equity
Nearly half of non-users, 47.1%, identified bias concerns as a barrier to adoption. That concern is appropriate. AI systems can reproduce or amplify inequities in the data they were trained on.
Accuracy and Reliability
Among non-users, 69.1% cited accuracy or reliability concerns. Psychologists should assume AI output requires verification, especially when it relates to clinical care, assessment, diagnosis, or documentation.
Professional Judgment
AI should support professional judgment, not replace it. The psychologist remains responsible for the quality, accuracy, and ethical appropriateness of the final work product.
What This Means for the Field
The study points to a clear conclusion: psychology needs more than general AI awareness.
The field needs practical, specific guidance that helps psychologists decide which AI uses are appropriate, which require safeguards, and which should be avoided.
That guidance should address:
Disclosure
Informed consent
HIPAA compliance
Business associate agreements
Documentation
Bias review
Human oversight
Assessment and report-writing boundaries
Training expectations
Clinic and agency policy development
The strongest message from the study is not that psychologists should trust AI more. It is that psychologists need better structures for using AI responsibly when they choose to use it.
Final Takeaways
56.9% of surveyed health service psychologists reported work-related AI use in the past six months.
AI users most often reported using AI for work-related questions, plain-language explanations, presentation content, and documentation support.
Only 20.0% of respondents said they trust AI, but 54.8% said it is helpful in practice.
Ethical concerns were strongest for test interpretation and clinical report writing.
Disclosure and privacy practices remain inconsistent and need clearer professional guidance.
The high level of interest in training suggests psychologists are ready for practical, ethics-grounded professional development.
Internal Links
For more context, see:
Earlier post: AI Use Among Health Service Psychologists: Current Use and Changes from 2024 to 2025
Preprint: Helpful but Not Trusted: Artificial Intelligence Adoption and Ethical Concerns Among Health Service Psychologists
https://osf.io/preprints/psyarxiv/gf9sq_v1
AI Use Consent Forms
https://lockwoodconsulting.net/blog/new-ai-consent-documents-for-psychologists-now-available
Interactive Checklist to Help with Adopting AI in Practice
https://lockwoodconsulting.net/blog/a-practical-checklist-for-evaluating-ai-psychology-tools
Call to Action
AI is already changing psychological practice. The next step is making sure that change is ethical, transparent, and grounded in professional judgment.
For more posts on AI, psychology, education, and responsible implementation, visit the blog and sign up for updates:
https://lockwoodconsulting.net/blog
You can also follow Lockwood Educational & Psychological Consulting on Facebook:
https://www.facebook.com/profile.php?id=61559517393918
AI Use Disclosure
AI Use Disclosure - Portions of this post were drafted with the assistance of an AI writing tool and revised by the author for accuracy, clarity, and professional judgment.