Legislating Innovation: A Scoping Review of the Scientific Evidence Underlying Illinois’ AI Mental Health Ban

Background and Purpose
In 2025, Illinois passed the first law in the United States prohibiting AI systems from providing psychotherapy or making therapeutic decisions without oversight from a licensed professional. As states regulate emerging technologies, these policies raise questions about democratic principles, including equitable access for marginalized populations who might choose AI tools due to cost and availability.

As we seek to harness technology for social good, it is important to interrogate such laws and their assumptions. Some see this legislation as protecting clients or the profession; others argue it prematurely restricts innovation and limits AI's potential to address unmet needs. These tensions raise key questions about what AI can and cannot do in mental health care, and the actual needs and preferences of professionals and consumers.

This study has two aims: (1) to identify assumptions about mental health services embedded in the Illinois law, and (2) to review the scholarly literature related to those assumptions, with attention to social work scholarship.

Methods
We conducted a content analysis of the Illinois legislation to identify key definitions, concepts, and assumptions related to mental health services, service delivery, and technology use.

We then conducted a scoping review to examine how these concepts are represented in the literature. Using university databases and AI-assisted search engines, we performed a systematic search using terms derived from the legislative concepts. Articles were reviewed for how scholars defined and discussed these concepts, with attention to relevance for social work research and practice. We excluded articles focused solely on AI as a technology without addressing psychosocial services or clinical practice.

Data were organized into a matrix in which each row represented an article (N = 108) and each column a concept from the legislation (N = 32), marking the presence of each concept.

Results
The content analysis identified six core concepts: definitions of psychotherapy services and therapeutic communication, permitted and prohibited AI use, consumer protection, and legal exceptions (e.g., peer-support, religious counseling). The scoping review identified 511 articles; after removing duplicates and applying exclusion criteria, 108 articles were analyzed.

We identified 32 categories, synthesized with legislative concepts into four themes: efficacy of agentic AI, AI-assisted therapy, consumer protection and accountability, and limitations of the law. Nine categories were not reflected in the legislation but were frequent in the literature.

Conclusions and Implications
We found strong support for three legislative concepts and challenges for three. While framed as protecting consumers, this law reinforces an existing boundary without clear evidence it reflects how people access care.

Restricting AI as a standalone option may increase risk by pushing consumers toward unregulated supports. This is concerning given that individuals use AI mental health tools not out of preference over human therapists, but because they are available and affordable.

As similar legislation spreads, many may soon live in states where standalone AI mental health services are illegal. These restrictions may limit access to care and social workers' use of AI tools, raising critical questions about whether such policies advance or undermine equitable access.