When I first arrived in Clearwater Basin, Alberta—population 12,000—I expected to find a small-town controversy about energy costs. What I discovered instead was the frontline of AI-generated climate misinformation that nearly derailed a decade-long environmental initiative.
“We thought it was coming from concerned citizens,” Mayor Claudia Vernier told me over coffee at the town’s main diner. She slid a stack of printed emails across the table. “These looked like they came from our neighbors, people we’ve known for years.”
The emails, nearly 300 sent to council members over two weeks, contained remarkably similar language opposing the town’s Climate Resilience Program—a modest initiative that had helped reduce municipal energy costs by 15% since 2015. The program was up for its scheduled five-year renewal vote.
What made town officials suspicious wasn’t the opposition itself but its sudden intensity and coordination. The messages contained identical data points about “hidden costs” that weren’t in any public documents.
My investigation began by tracing IP addresses from the emails. Working with digital forensics expert Dr. Maya Chandrasekaran from the University of Alberta’s Digital Ethics Lab, we discovered that 87% of the messages originated from temporary email accounts created through VPN services.
“This is a textbook case of synthetic grassroots pressure,” Chandrasekaran explained. “The sophistication suggests AI-generated content customized to mimic local concerns.”
The campaign centered on a document titled “Clearwater Basin Economic Impact Assessment” that circulated widely on Facebook and via text messages. The 32-page report included convincing graphs, testimonials from “residents,” and economic projections showing the climate program would raise household costs by $3,200 annually.
Court records I obtained show the document was initially distributed through five Facebook accounts created just weeks before the campaign began. These accounts claimed to belong to local business owners—none of whom actually existed.
“I thought I was reading something from our chamber of commerce,” said Frank Doherty, a third-generation hardware store owner who initially opposed the program renewal based on the report. “It referenced local landmarks, mentioned our specific industries. It looked legitimate.”
The report’s apparent source was the “Kiclei Institute for Economic Freedom”—an organization that proved impossible to locate. The name itself was a clue: provincial records expert Teresa Blackfoot pointed out that “Kiclei” is an anagram for “I like CI”—with CI potentially standing for “climate initiative.”
“This is next-level disinformation,” said Dr. Alan Rosenthal, director of the Climate Misinformation Project at the University of Toronto. “We’re seeing AI tools used to create hyperlocal campaigns targeting specific communities where a single vote can change environmental policy.”
The sophistication of the campaign went beyond fake documents. Residents reported receiving personalized text messages referencing their businesses, children’s activities, and specific concerns about the local economy.
I reviewed voice messages left for three council members where the callers—supposedly local residents—discussed neighborhood-specific concerns. Voice analysis conducted by Citizen Lab showed these were likely AI-generated using publicly available council meeting recordings to mimic regional accents.
“The technology to create this kind of personalized misinformation campaign is widely available,” explained Elena Westbrook of the Electronic Frontier Foundation. “For about $200, someone could generate hundreds of customized messages that reference local details harvested from public social media.”
The climate program renewal vote failed by a single vote in May. However, after my initial reporting revealed the misinformation campaign, council member Douglas Reid requested a revote.
“I was influenced by what I believed were my constituents,” Reid admitted during the council meeting. “Finding out these weren’t real people but computer-generated messages made me question my decision.”
The Alberta RCMP’s Cybercrime Investigation Unit has opened an investigation, but identifying the campaign’s source faces significant challenges. Detective Sergeant William Chen told me they’re examining potential connections to similar campaigns in three other Canadian municipalities with climate initiatives.
“We’re seeing a pattern where small-town environmental policies face coordinated opposition that disappears once the vote happens,” Chen said. “The perpetrators are sophisticated enough to cover their tracks.”
Federal records show that Clearwater Basin’s Climate Resilience Program had reduced municipal energy costs by $382,000 annually through building efficiency upgrades and small-scale renewable projects—savings that helped fund the town’s new emergency services building.
The program passed on its second vote last week, but the incident has left residents shaken. Town council has now established a verification system for public comments, requiring confirmation of identity for official feedback.
“What scares me isn’t just that we almost lost a beneficial program,” Mayor Vernier said during the revote meeting. “It’s that we can no longer trust that what we’re reading or hearing is coming from actual neighbors.”
As AI tools become more accessible, experts warn that small communities with limited technical resources are particularly vulnerable to such campaigns. The Climate Action Network Canada has documented similar patterns in twelve municipalities across four provinces since January.
“When misinformation is tailored to your town, mentions your streets and businesses, and appears to come from people you know, even the most discerning citizens can be fooled,” Rosenthal noted. “That’s what makes this new frontier of AI-generated campaigns so effective—and so dangerous.”
For Clearwater Basin, the episode has transformed a routine policy renewal into a cautionary tale about democracy in the age of artificial intelligence.
