AI-powered emails are influencing politics on hot-button issues

Two weeks before the board of a Southern California environmental agency was to vote on a pair of landmark climate initiatives last May, staff raised an alarm. "An aggressive campaign" from an "AI powered platform," they told executives, was flooding their servers with thousands of emails from the public opposing the proposals to phase out gas water heaters and furnaces in nearly half the state's homes.

Officials at the South Coast Air Quality Management District scrambled to figure out who was behind the deluge - and if the emails were genuine, according to documents obtained by Bloomberg News in a public records request.

One message sent to an air district administrator came from a friend who said he didn't write it. Another purported to be from the agency's own executive officer, Wayne Nastri. (He later got an email from a group fighting the proposals, thanking him "for voicing your opposition to these costly rules" - rules that his staff had drafted.)

Nastri's deputy Lisa Tanaka O'Malley floated a theory: "I'm wondering if they used AI to scan for emails and sent a massive blast out." An attempt by the agency to verify the emails proved inconclusive: Officials contacted 172 commenters but only five replied, with two confirming they'd sent the letters and three denying they had done so.

Ultimately, the air district's board voted seven to five against the proposals put forward by its staff.

From California to New York to North Carolina, recent plans for new energy projects or green regulations have spurred unusually high volumes of public comments, confounding local and state officials. Originating from paid campaigns and generated with the help of automation tools and artificial intelligence, the comments have almost universally expressed support for fossil fuels or opposition to restricting their use, as well as backing other policies favored by corporate interests.

It's not entirely clear in what ways these campaigns used AI or how much they swayed policymakers. But their mere existence has led to fears about AI being harnessed to misrepresent public opinion on issues that are already contentious and politically polarized, and has prompted discussion of how to guard against bad actors.

"Public comments are integral to the functioning of democracy," said Samuel Woolley, a disinformation expert at the University of Pittsburgh. "They have been since ancient Athens." When ordinary citizens can weigh in on development plans and policies, government officials are pushed to be more transparent and engage with a range of opinions.

Yet the very openness of the process leaves it vulnerable to being co-opted. Especially in the age of AI.

South Coast air district spokesperson Connie Villanueva said the agency is studying ways to screen out potentially AI-generated comments, including creating a portal requiring additional human verification. "Maintaining the integrity and transparency of the public comment process is a top priority for the agency," she said in an email.

Activist groups, including the Sierra Club, have asked the agency to launch an investigation into the campaign from last year. "It's extremely concerning because if elected officials and regulators can't trust the comments that they're receiving, that undermines the public decision-making process," said Dylan Plummer, a senior campaign advisor at the Sierra Club.

Villanueva said the air district can't comment on any possible investigation.

"If elected officials and regulators can't trust the comments that they're receiving, that undermines the public decision-making process"

CiviClick, the Washington, DC-based firm behind the emails to the agency, promises clients it can deploy AI to instantly create uniquely personalized letters from advocates "at the scale of a mass campaign." But not all clients opt to use this function. The political consultants who ran the South Coast campaign did not, according to CiviClick.

"To be precise: AI was not used," the company wrote in an online post. "No artificial intelligence, no machine learning, and no automated text generation touched the messages submitted in that workflow."

Nevertheless, critics say AI is shaping up to be a new tool for an old problem: astroturfing, when companies or interest groups create campaigns designed to look as if they started on the ground. Automation has already amplified such "fake grassroots" efforts, Woolley said. He added, "So to use generative AI in this kind of context just puts everything on steroids."

CiviClick was also behind a New York effort last year that sent nearly 10,000 emails on behalf of pipeline union members backing the continued use of natural gas in a state energy plan, which ended up being approved. Roughly half the letters, however, were submitted after the public comment period deadline had passed, based on materials obtained via a public records request.

That missed deadline isn't the only bump that CiviClick-linked campaigns have experienced. In Southern California, the company blasted more than 12,000 messages to the air district over a five-day period, but nearly 10,000 of them ended up in a spam filter, according to agency records. In North Carolina, the energy company Williams Cos. used CiviClick's software to convince officials of popular support for extending its gas pipeline, but officials were immediately skeptical.

"Something looked off" about the high volume of messages, said Forsyth County Commissioner Dan Besse, and so did the fact that they "were coming at all times of day and night." Besse added that at least three people had told the commission they never sent the comment attributed to them, calling that "a huge red flag."

Asked about such suspicions, Chazz Clevinger, the chief executive officer of CiviClick, said the company never fakes emails or identities. The idea that messages in North Carolina or California "were systematically generated, fabricated or submitted without the participation of real people," he said, "is patently false."

Clevinger offered explanations for why some people said they hadn't written the emails under their names. They could have forgotten, submitted on behalf of a family member or given the wrong information because of shared web cookies. The company tries to prevent impersonation, he said, and its rate of "after-the-fact disputes," as he termed them, is in line with those of letter-writing drives and call-in campaigns.

Williams Cos. also denied that letters tied to its campaign were disingenuous. "It is possible in any high-volume campaign that a handful of letter-senders will not remember hitting ‘send' on the email, but we have a time stamp and IP record confirming when each email was sent," said company spokesperson Cherice Corley.

"If elected officials and regulators can't trust the comments that they're receiving, that undermines the public decision-making process"

Not surprisingly, news coverage of these campaigns alarmed policymakers elsewhere. "We need to put defenses in place!" Rich Chien, a manager at the Bay Area Air District, a pollution regulator, wrote in a February email to a colleague after the Los Angeles Times revealed the Southern California campaign.

By that time, Speak4, another digital advocacy outfit that touts its use of AI to "personalize messages at scale," was already sending emails to Chien's agency. The messages supported a proposed socio-economic impact study of air regulations, which environmentalists said could deter tightening them. Several people whose names were on the emails told the San Francisco Chronicle that they never sent them.

Speak4 CEO Joe Mansour said letter-writers submitted their comments by filling out a web form. "No AI was used in the campaign, and Speak4's AI tools were not used or made available for the campaign," he said in an email. Regarding people who said they hadn't written letters, Mansour said, "Speak4 did not write or submit messages on behalf of those individuals."

Bay Area Air District spokesperson Tina Landis said a new system will soon be rolled out that lets the public directly submit comments through the agency's website to "help address concerns about automated email submissions."

Officials in California, New York and North Carolina were able to detect the mass email campaigns since they came from identical email addresses. The letters themselves also resembled those from past advocacy pushes - they read like conventional form letters, not AI creations. But that might not last very long, if a recent campaign in Berkeley, California, is any indication.

The texts from a group called the East Bay Alliance for Public Safety multiplied as a bitter fight over a Berkeley Police Department proposal to deploy surveillance cameras and drones in the liberal stronghold approached a May vote by the city council. The $2 million contract would expand an existing relationship with Flock Safety, an Atlanta-based company that operates in cities across the U.S.

"Tell Berkeley leaders to support safer neighborhoods. Answer one quick question to generate a personalized message and make your voice heard," read one text that appeared on the phone of a Bloomberg News reporter who lives in Berkeley.

Clicking on the link in the text took you to a website operated by Influent, a Los Angeles digital advocacy firm that "offers personalized data collection tools that leverage the power of AI to help individuals craft thoughtful, personal materials."

The site first inquired, "What's your position on safety cameras in Berkeley?" If the answer was "oppose," you were thanked for "participating in our poll!" Reply "support," though, and the site then asked for an anecdote about being affected by crime, along with your contact information. If you then pressed "send," seconds later a draft of a personalized email containing the anecdote - addressed to the mayor and city council - appeared in your inbox for review. Then you, not Influent, sent the email to the officials.

We entered an actual crime experience several separate times with the same wording. Each letter Influent returned for review employed different language and syntax to recount the anecdote and to make the argument for surveillance cameras. The messages also varied in length and subject lines. Each appeared, more or less, to be "original," suggesting the use of generative AI.

"It's frustrating when crime impacts your life directly," read one email, putting words in the writer's mouth, while another declared, "This kind of thing makes me worry about safety in our city."

In contrast, a campaign run by progressive digital advocacy firm Action Network sent dozens of nearly identical emails from Berkeley residents opposing the cameras, shows a review of documents obtained in a public records request. The campaign was easily identifiable as the messages were sent from Action Network's email address.

Not so the Influent campaign. "We were not aware of this campaign and, as far as we can tell, did not receive" messages from it, Melissa Male, communications director for Berkeley Mayor Adena Ishii, said in an email.

Neither Influent nor the East Bay Alliance responded to requests for interviews.

Plummer, of the Sierra Club, said that AI-assisted campaigns "seem like an effort to mislead public officials to make them believe that these comments were written by individuals who cared enough to write out a thoughtful, unique comment."

Berkeley City Council member Brent Blackaby recalled receiving emails like those generated by Influent that weaved in personal anecdotes in support of the cameras. But he said written comments, including personal stories, are just one factor that officials consider when weighing a policy decision. Blackaby, who formerly worked for a digital advocacy agency, said he's not particularly concerned about such AI-powered campaigns as long as residents authorized the messages.

"It does add to the challenge for policymakers to determine how representative this is of the people that we are being elected to serve," he said.

In the end, it was old-fashioned IRL - in real life - advocacy that appeared to carry the day in Berkeley. Camera opponents campaigned door-to-door and showed up by the hundreds at city council meetings. In May, the majority of the city council, including Blackaby, that had previously leaned in favor of the Flock contract expansion voted to kill it.

____

-With assistance from Jordi Ng.

Copyright 2026 Tribune Content Agency. All Rights Reserved.

This story was originally published June 27, 2026 at 9:04 AM.