AI Analysis of Public Comments in Rulemaking

Rulemaking can generate large public records containing unique comments, form letters, technical attachments, and repeated text. AI tools can help sort comments by topic, identify similar submissions, and create summaries that direct analysts to relevant parts of the record. Similar wording can arise from organized advocacy, shared templates, or independent agreement; it does not establish identity or authenticity. A system may also miss distinct arguments expressed in different language, over-compress a minority view, or misread an attachment. Agencies should preserve submissions as received, keep the method used to classify or deduplicate them, and provide a path from each summary back to the underlying record. Any conclusion about the rule should reflect relevant comments and supporting evidence, not merely frequency. The Administrative Procedure Act and agency-specific procedures govern rulemaking duties; the guide does not provide legal interpretation for a specific proceeding. A defensible workflow documents how automated analysis was used, what human review occurred, and how material issues were addressed. Agencies should distinguish a public submission from a verified person or organization and avoid dismissing comments based solely on automated authenticity estimates. Data privacy, accessibility, and public-record retention also matter. AI can help staff navigate a large record, but it cannot decide which comments matter legally or replace the agency’s reasoned decision-making. Reviewers should verify quotations, preserve context, and ensure the final record supports the agency’s explanation.