Purpose In May of 2024, the U.S. Drug Enforcement Agency (DEA) proposed to reschedule marijuana from Schedule I to Schedule III of the Controlled Substances Act (CSA). As part of the rescheduling process, public comments were solicited online during a three-month period. At the end of the public comment period, 42,903 comments were submitted. The collection and analysis of a large volume of public comments is a challenge for policymakers, especially when the timeline for policymaking is finite. The purpose of this study is to respond to emerging scholarship that has identified the need for transformative changes to make planning and policymaking processes more collaborative and effective using advanced tools. Design/methodology/approach This study proposes a methodology for using ChatGPT to perform sentiment analysis and topic modeling on public comment data as a tool to enhance the citizen participation process. This analysis examines the sentiments in public comments submitted to the DEA about marijuana rescheduling and highlights professional and institutional tendencies voiced in them. Findings This study compared sentiments expressed in unique comments (43.6%) to duplicate comments written by institutional interests and posted by their constituents (46.3%) and found that duplicate comments were less likely to agree with rescheduling marijuana and more likely to advocate for descheduling marijuana from the CSA entirely. Practical implications The use of ChatGPT to analyze public comment data has general applicability across the rulemaking process and allows policymakers to systematically analyze the opinions of the constituency. This analysis provides policymakers with evidence to support efforts to build artificial intelligence (AI) capacity and put data protection systems in place to maintain the trust of participants in public deliberations. Originality/value This study introduces methods to apply AI tools such as ChatGPT in citizen participation. These methods can increase the capacity of government agencies to analyze volumes of data that come from the public participation process.
Yin et al. (2026) studied this question.