Campaign season is taking shape across Ontario, bringing with it the usual mix of signs, slogans, flyers, and social media posts. But this time, there is another tool on the campaign trail: Artificial intelligence. In the town of LaSalle, Ont., all three candidates running for mayor tell CTV News they are using generative AI in some form — whether to create graphics, research issues, or take care of smaller editing tasks. Where they differ is what they are willing to let it do. Three candidates, three boundaries Incumbent Mayor Crystal Meloche said she has used AI to help create graphics for her campaign, working from branding, colours, and messaging developed with her daughter. But she said the substance behind those graphics remains her own. “It’s a great tool, but I want to make sure everything else is authentically me,” Meloche said. She said her platform comes from her experience on council and conversations with residents, not from a chatbot. “It’s my words 100 per cent,” she said. “I think AI can’t replace that, right? That’s my knowledge. That’s my information, and that’s what I share with people.” Meloche said she is opposed to using AI to create fake videos or voiceovers, drawing a distinction between using the technology as a design tool and allowing it to become the candidate. Michael Macri draws the line somewhere else. He said his brochures, website, colour scheme and campaign messaging are his own, but he uses AI extensively behind the scenes for research, summarization, and analysis. Macri described it as something close to a research assistant or sounding board — including asking it to argue against him. “I ask for a different perspective,” Macri said. “A lot of times when I’m using artificial intelligence, I’ll say my position is X, Y, Z. How would you counter that? And then I want to hear the counter argument to make sure that I don’t have any blind spots or that I’m being fair.” He said he also uses AI to work through municipal issues, provincial legislation, and long-term financial modelling. But Macri cautioned against treating its answers as gospel. “It can be wrong and it can be wrong a lot,” he said, adding he will sometimes ask the same question seven different ways to see whether he gets a consistent answer. Shaun Dass, meanwhile, said at least 90 per cent of his campaign material is generated by people, including graphic designers and web designers. “But what we’ve learned in academia is to incorporate it,” Dass said. He acknowledged using AI for some editing and smaller tasks. “Of course, we do use AI for some editing and things like that. Some minor stuff, some tedious tasks,” he said. Dass also said he generated a poster for one of his upcoming coffee chats. But asked whether he would use AI to answer questions from residents, his answer was emphatic. “Heck no. Heck no, right? That is not something I’m about,” Dass said. “It’s always fresh. It’s always original. And if it’s an online question or if it’s an email in question, never.” A campaign tool that wasn’t here last time Ontario’s last municipal election was held on Oct. 24, 2022. ChatGPT launched publicly just over a month later, on Nov. 30, 2022. Four years on, candidates are heading into the first Ontario municipal election with consumer generative AI tools widely available — and substantially more capable than they were even a few years ago. York University public policy professor Zac Spicer said that makes this election something of a first. “This is the first kind of a local election I think we’ve had here,” Spicer said, noting how quickly commercially available AI systems have advanced since 2022. “All these retail AI systems have taken leaps and bounds since. You can make really quality graphics without even a subscription to even the basic system like ChatGPT.” A campaign in seconds Generative AI has dramatically lowered the technical barrier to creating campaign material. Instead of starting with design software, a blank page, or a hired professional, a user can describe what they want in plain language. From that description, tools such as ChatGPT or Google Gemini can generate written copy, slogans, and increasingly sophisticated images. The person using the tool still decides what to ask for, what to keep, and what to publish — but much of the work that once required specialized writing or design skills can now be produced almost instantly. That speed works in both directions. The same technology that can create a polished campaign graphic can also generate negative messaging, misleading material or content designed to imitate the look and feel of a real campaign. For Dave Cormier, director of curriculum development and delivery with Open Learning at Thompson Rivers University in B.C., that ease is part of the concern. “It’s a 10 second operation,” Cormier said. “So, you can go to any one of these tools. Something like ChatGPT or chat, as the kids call it. And just give it an instruction and say, make me a poster that says this.” Cormier said the bigger issue is not simply whether AI can produce convincing campaign material, but how quickly false or misleading information can be created alongside it. “Misinformation is probably the most important,” he said. “At this point, anybody can generate anything very, very quickly. It doesn’t need to be professional. It doesn’t need to be nice. It doesn’t need to be true.” Cheap, quick — and attractive to campaigns There is, however, a practical reason candidates may be reaching for the technology. Municipal campaigns are not known for bottomless budgets. Meloche acknowledged AI-assisted design could mean less work going to professionals, but said candidates have spending limits and choices to make about where their campaign dollars go. “I mean, obviously it takes business away from people. So, there is that negative part of it,” Meloche said. “But I’m on a budget like everyone else. When you run an election, they give you an amount of money you’re allowed to spend. And so, we have to decide what we’re going to spend it on.” She said she chose to direct more of her available money toward physical signs and flyers. Macri sees AI as a productivity tool. “I think AI is an incredible productivity tool,” he said. “And it can strengthen your strengths and compensate where you may have an opportunity.” He also acknowledged the disruption the technology could bring to creative industries. “It’s going to happen where you had seven people doing creative. Maybe now you only need two,” Macri said. “And so that’s going to happen. And I feel horrible for those five people.” But he argued new jobs will emerge as others disappear. “As much as it’s going to take jobs away, it’s going to create other jobs,” he said. Spicer said the financial pressure is particularly relevant in local politics, where candidates are expected to produce increasingly polished content without necessarily having the money or staff of a provincial or federal campaign. “A lot of candidates aren’t working with a ton of money, and there is this sort of expectation that you have to have a fairly polished video or graphics package,” he said. In the past, he said that could mean hiring a graphic designer or finding a volunteer who knew how to use professional design software. “You could create that now very quickly,” Spicer said. Cormier accepts the accessibility argument — to a point. “I understand people’s frustration, that it can be expensive to do these things. And certainly, being an election can be really expensive as well,” he said. But he argues that something can be lost when professional work becomes a prompt away. “Any time we take something that is a professional activity, and we make it simple to give people accessibility, we lose the professionalism of that activity,” Cormier said. The problem of trust For Cormier, one possible response is simple: Disclose it. “I absolutely think anybody who uses AI for anything should declare that they’ve done it,” he said. “I think that transparency is, again, something we want in our elected officials.” He said voters should also go back to the source when something online looks suspicious — checking a candidate’s official accounts and, where possible, speaking to the candidate directly. Dass offered similar advice. “Don’t just live off your social media feed,” he said. “Go and see them, meet them, talk to them, watch or listen to their interviews, wherever they might be.” For University of Windsor political science department head Lydia Miljan, the test is ultimately more old-fashioned: Does the person voters meet match the candidate being sold online? AI, she said, “certainly can provide a slick product.” “But at the end of the day, when you come to my door and you can’t string a sentence together, I’m going to judge you accordingly,” Miljan said. “The AI product has got to align with what the actual politician is selling.” Her concern goes beyond campaign graphics. If candidates begin handing more of the intellectual work over to AI, she said, voters may eventually have to ask a more fundamental question. “Why are we electing people to go to office if they can’t do the work that’s required of them?” Who draws the line? Spicer said candidates should also be mindful that voters may view AI-generated material differently than the people creating it expect. He pointed to research suggesting viewers can perceive AI-generated images and video as inauthentic rather than simply cheaper or easier. “People, for the most part, see it as an intention to deceive,” Spicer said. That does not make Spicer an AI pessimist. He said the technology can make campaigning more accessible and efficient. But the boundary may not ultimately be set by candidates, academics, or the technology companies building the tools. It may be set at the ballot box. “I think the boundary is probably going to be defined by voters who will say, we don’t like this or we want to hear from you, not the AI generated version of you,” Spicer said. AI is already part of the campaign. How much of the candidate it replaces — or reveals — may now be up to voters.
