June 4, 2026
Artificial intelligence (AI) chatbots are under scrutiny for the way they respond to users' private thoughts. A new use case for the technology has turned the focus to more public conversations; specifically, the way AI-generated feedback can game the public comment system.
Earlier this spring, journalists, advocates, and public officials concluded that an AI platform may have helped generate 20,000 emails that flooded the South Coast Air Quality Management District (AQMD), ultimately swaying a decision by board members last June to resist a rule that would phase out gas-powered appliances in Southern California.
The 7-5 decision — which many believe was affected by large-language models and agentic AI, autonomous artificial intelligence that can complete multi-step processes — has been a catalyst for commentary and angst about new technology, with public officials wrestling with how to avoid letting nefarious actors or technology ruin a process central to urban planning and regulatory reviews.
"During COVID, when meetings had to go online, hybrid commenting became a de facto practice," says Kelly Hickler, AICP, founder of Cinnovation Studio and co-chair of the American Planning Association's (APA) Community Engagement Interest Group. "By making the process available online, you are widening the scope of people whom you're now able to communicate to and get feedback from: those with caregiving responsibilities, mobility issues, shift work jobs, or disabilities. You can easily get hundreds or thousands of legitimate online comments for projects people really care about."
The problem now is differentiating between those legitimate comments and ones generated by AI.
"As a mayor or a senator, how do I know what really matters to the public anymore?"
—Christopher Cabaldon, California state senator
Gaming the public comment system isn't a new phenomenon, says Nader Afzalan, founder and CEO of The Triangle and a University of California-Berkeley professor who focuses in part on the intersections of technology and community feedback. These new technologies make it even more vital to understand the accurate views of the public.
"I think there should be a way to validate the engagement; not just that it's a person, but who you are and what type of community you're representing," says Afzalan, who is also a member of APA's Technology Division.
The debate was the impetus for a new Senate bill (SB), introduced February 18 by California State Senator Christopher Cabaldon — SB 1159 — that would allow public agencies to ignore feedback generated by AI. For the purposes of open meeting and public records statutes, the law would clarify that "person" and "member of the public" do not include AI systems, autonomous agents, robots, or other nonhuman entities.
"Part of the underlying challenge here is that we all have opinions, and the public engagement system for government is set up so that there's some friction, some effort, needed to demonstrate your opinion," says Cabaldon, a former mayor of West Sacramento. "The extent to which AI can make that effort zero is a challenge to the government, because as a mayor or a senator, how do I know what really matters to the public anymore?"
A problem that's spreading
This issue isn't going away. A series of similar emails in support of a gas pipeline in North Carolina led county commissioners to suspect a larger campaign was in the works in September 2025. Around that time, news emerged in the U.K. about an AI-powered service that allows a citizen to register their objection to housing projects, for example, with the system automatically sending their comment to every project under review.
In Northern California, there were allegations of AI tools fabricating comments from real residents who say they never offered feedback about pollution regulations under review by the Bay Area Air Quality Management District. This led groups like the Sierra Club to call for an investigation by the state's attorney general's office.
The potential for these technologies to mislead public officials and undermine real people's engagement in regulatory matters is concerning, says Dylan Plummer, a Sierra Club campaign advisor. "Grassroots engagement on these public processes can really shape how they turn out," he says. "It's a double-edged sword, because the more AI comments are submitted, the less credible all comments in general will become and, as public agencies start putting in more guardrails, it'll have a discouraging effect on people engaging in these processes. It feels like a race to the bottom where we're still losing and creating barriers to public engagement."
Protecting the integrity of public comments
The rapid development of AI, and the potential to use technology to recast public opinion — and with it, regulatory and political decisions — creates a new challenge for planners. Software that uses AI to create advocacy tools and services that can help users generate commentary that matches their viewpoint should, in theory, help sharpen their argument to gain the attention of public officials. However, this kind of technology could also be co-opted to create and generate reams of fake, low-value commentary that can overwhelm a public forum.
A South Coast AQMD spokesperson tells Planning that they determined the email submissions were suspect after an internal review process that parsed language patterns and metadata, and then contracted a small subset of roughly 175 emailers to confirm whether they had sent a comment. Three of the five who responded said they had not.
While South Coast AQMD didn't definitively conclude the comments were AI-generated, it did raise the alarm. The agency is now working on refining its comment process. One option being considered is implementing a system that requires human verification for all submissions.
"It's a double-edged sword, because the more AI comments are submitted, the less credible all comments in general will become."— Dylan Plummer, campaign advisor for the Sierra Club
"Maintaining the integrity of our public process is a top priority," according to South Coast AQMD in a statement provided to Planning. "When a firm or campaign may be retained and funded to generate and submit large volumes of comments as part of an organized outreach effort, which can include structured email campaigns that facilitate or automate submissions on behalf of individuals, it can be difficult to verify. Given the evolving landscape of AI, we are looking into ways to receive public comments electronically that ensure integrity while still maintaining a public process that encourages all stakeholders to participate and provide comments."
In the future, public officials might have to figure out methods for soliciting public feedback that encourage participation and yield relevant, valuable insights. Afzalan suggests asking for a few pieces of validating data , such as a ZIP code or a question that establishes the participant's stake in the issue. For example, a referendum about bike lanes might ask if someone is a regular cyclist. Another way might be to grant anonymity after someone is validated to encourage more upfront responses.
"Engagement should be deliberative," Afzalan says. "If it's not, it's just a random engagement."
There's a delicate balance in getting people to comment online if more exhaustive and potentially intrusive data collection and verification is required, says Hickler, who co-authored a new guide about hybrid commenting best practices. One fear is data leakage, which can discourage some potential commenters from interacting. Hickler suggests using small means of verification, such as asking for ZIP codes, to make it easy for residents to comment without creating additional privacy bottlenecks.
As the use of AI to sift through public comments becomes more common, California, for example, is using it to gather feedback about how the technology is affecting jobs and the workplace, it's important that any rules or regulations recognize the complexity of the issue. Cabaldon stresses that his Senate bill seeks to make it clear that, for the purposes of public commentary, the actual public is the only one with a voice.
"How do we design a modern civic engagement system that isn't just defending against technological change but is also recognizing that change?" Cabaldon asks. "For a lot of the advocacy that happens in the U.S., it's essential."
RECOMMENDED ARTICLES
