TORONTO — Eric Apps had an inkling of who would get his vote in Toronto’s municipal election, but the lawyer wanted a second opinion before making his final decision. In Parkdale-High Park, his usual safe choice in longtime councillor Gord Perks was taken away when the 63-year-old decided to retire. The race for the city’s west-end ward has ballooned to 10 candidates — none of whom Apps is familiar with. He turned to a system he was familiar with: Claude, Anthropic’s artificial intelligence chatbot. Apps asked the bot for the top five polling candidates in his ward and their policies on issues such as bike lanes, housing and property taxes. “It laid them all out in a very easily digestible format,” he said. Apps is not the only one turning to AI for election information. During Quebec’s provincial election earlier this month, an estimated 40 per cent of the province’s voters turned to AI chatbots and received false and outdated information about the election, according to McGill University researchers. Many Quebec news outlets don’t allow chatbots to access their websites, so their reports cannot be mined for users’ queries. Experts say this marks a troubling shift in the democratic process, where citizen engagement and human judgment is superseded by some algorithms that could be providing false information. There were a few gaps in Apps’ own search, he noted. The chatbot could not provide much information on candidates who weren’t as widely reported on by news outlets or who didn’t have very active online presences. He acknowledged that some of the gaps could be related to the parameters of his questions. “The blind spots are probably in the eyes of the beholders,” Apps said. “If you have your point of view going into it, then it probably is going to be giving back information that is not necessarily going to focus on opening your eyes.” Alissa Centivany, an associate professor of Information and media studies at Western University, said artificial intelligence bots have replaced traditional search engines, such as Google, and reliable media platforms, such as news organizations’ websites. For some voters, chatbots have become their main method of seeking information. For AI researchers like Centivany, it remains unclear what an AI chatbot’s “technical process” is as it scrapes the internet for information. Centivany notes that the quality of information AI chatbots can compile is often lower than she would hope for, especially in the context of political elections. One concerning element, she said, is that the mishmash of information it gathers can include user-generated content from social media sites like Reddit. Centivany said researchers are starting to see evidence of how a bot will compile information based on what they think a user will want to hear, based on past queries, as opposed to objective facts. “A lot of it is opinion, a lot of it is untrue, a lot of it is inflammatory,” she said. “And it’s being used as a source of information.” As he asked Claude more questions about candidates in his ward, Apps said he was also keenly aware of its limitations, especially with the influx of content on social media platforms, such as TikTok and Instagram. “You have to take your AI engine with a grain of salt,” he said. “It gave me the basic information and then it puts it back to the viewer.” The accessibility of the AI platforms made political participation easier, Apps said. While he considers himself somewhat politically aware, he said the chatbot acted as an information “refresher.” “You ask it a question and it gives you an answer,” he said. “It’s not giving you glossy videos with clickbait-ish headlines.” Anthropic said Claude uses web search to find up-to-date information and shows its sourcing so users can “judge their credibility for themselves.” “Claude is trained to note when claims are disputed, when sources conflict, or when a source may be inaccurate or have an agenda,” the company said in a statement to The Canadian Press. Earlier this month, Anthropic announced new updates to its usage policies, which included a “refocusing” of their elections section and political content restrictions. “We’ve further sharpened our section on elections, focusing it specifically on disallowing the use of Claude to deceive voters or disrupt elections (for instance, by spreading false information about candidates or how to vote, impersonating candidates or election officials, or trying to suppress turnout),” the company’s news release read. Alex Marland, a professor of Canadian politics and the Jarislowsky Chair in Trust and Political Leadership at Acadia University, said it makes sense that voters would turn to an accessible tool such as AI for information about candidates in elections. Typically, people might vote for the political party they most align with, but candidates in Ontario municipal elections are not affiliated with a party. “What ends up happening is you have to invest a lot more time to try to understand, ‘What are my choices and what do these people stand for?’” Marland said. It’s a positive sign that voters want to get engaged in municipal politics, Marland said, since Ontario’s last municipal election had record-low voter turnout. According to the Association of Municipalities of Ontario, the average voter turnout in the 2022 was almost 33 per cent, compared to roughly 38 per cent in 2018. But it’s up to users to be vigilant in fact-checking the information they’re receiving, Marland added. “This is a piece of information that is going in with other pieces of information.” This report by The Canadian Press was first published Oct. 24, 2026. Sadeen Mohsen, The Canadian Press
