Support for Regulating Generative AI Mental Health Advice in New Jersey

• In Event: Artificial Intelligence and Governance

Abstract

Approximately one in six adults in the United States report using generative artificial intelligence (AI) chatbots (e.g., ChatGPT, Google Gemini, Claude) for mental health advice in the past year. There is evidence though, that AI chatbots may respond inappropriately to mental health-related prompts and multiple chatbots have been cited as direct or contributing factors in multiple suicides. As states consider a range of policies related to use of AI, we aimed to assess public support for a state policy regulating general-use AI chatbots’ interactions with users in the context of providing mental health advice. In March 2026, we fielded measures on use of generative AI for health advice and support for state policy regulating generative AI as part of the Rutgers-Eagleton/SSRS Garden State Panel Omnibus survey, an online, probability-based survey of adults (aged 18 and older) in New Jersey (N=1568). We assessed the proportion of individuals supporting regulation of AI chatbots in the context of mental health overall and by prior use of generative AI for mental health advice; differences in proportions were assessed using Wald tests. We also examined associations of individual characteristics (i.e., age, gender, race/ethnicity, education, income, employment status, marital status, disability status, region of state, political party affiliation, frequency of internet use) with policy support using logistic regression. All analyses used survey weights to represent the full New Jersey adult population. Most adults in New Jersey (59.5%) support a state policy regulating how generative AI chatbots interact with users seeking mental health advice. A greater proportion Individuals who have used an AI chatbot specifically for mental health advice were supportive (‘somewhat support’ or ‘strongly support’) of regulation compared to those who had not used AI for mental health advice (69.4% vs. 58.0%, p=0.011). The opposite was true for strong support for AI mental health regulation (23.4% have used AI vs. 32.6% have not used AI, p=0.02). In adjusted models for both overall and strong support, political party affiliation was the most significant predictor of policy support with both Independents (overall support odds ratio (OR): 0.51, 95% confidence interval (CI): 0.36-0.72; strong support OR: 0.44, 95% CI: 0.32-0.62) and Republicans (overall support OR: 0.25, 95%: 0.17-0.37; strong support OR: 0.19, 95% CI: 0.12-0.29) reported significantly less policy support compared to Democrats. Other demographic characteristics were not associated with policy support. Most adults in New Jersey support a state policy regulating how general-use AI chatbots interact with people seeking mental health advice. When assessing differences in policy support based on experience using AI chatbots for mental health advice, we identify two key groups supporting AI regulation: 1) those who may be skeptical of use of AI in the context of mental health (strongly support regulation and have not used AI for mental health advice) and 2) individuals who have used AI for mental health advice and support regulation of AI in this context. Understanding the concerns of these populations can help inform effective policymaking in the context of AI use for mental health advice.

Author

• Elizabeth M Stone, Rutgers University
  Presenting Author