Imagine standing in line at a community centre on election day, phone in hand, asking a chatbot a simple question: “Where do I vote?” The answer feels confident, friendly, and immediate. You follow it to the wrong address, miss your window, and later learn the chatbot had invented the location entirely. This is not a dystopian fantasy. It is the kind of everyday confusion now spreading through Canadian elections as artificial intelligence tools become more common in political life. For many people, AI chatbots are like helpful neighbours who always have an answer—except sometimes they have no idea what they are talking about. Researchers studying the impact of these tools are increasingly concerned that voters are being misled, not by malicious humans necessarily, but by the very technology designed to make life easier. The problem is not just that AI makes mistakes. It is that the mistakes look so trustworthy, so polished, and so human that ordinary Canadians may have no reason to doubt them. That is why researchers and advocacy groups are now urging lawmakers to bring AI chatbots under the clear authority of Canada’s election laws, before the next federal vote turns into a minefield of automated misinformation.
The research emerging from academic and civic organizations paints a troubling picture. When asked about basic voting information—polling hours, voter ID requirements, candidate positions, even the date of the election itself—chatbots frequently produce plausible-sounding but false answers. In one study, researchers posed the same set of questions to multiple AI systems and found that responses varied wildly, with some bots confidently stating that identification rules had changed when they had not, or inventing candidate biographies that bore no resemblance to reality. What makes this especially dangerous is the way AI composes language. It does not say, “I’m not sure.” It says, “You will need your driver’s licence and a utility bill,” even if that is not the law. It speaks in full, grammatically perfect sentences, with a tone of calm authority. For a voter who is already anxious about the complexities of civic participation, that voice can be incredibly persuasive. Researchers note that this is not a glitch that can be easily fixed by better coding. The very design of large language models—which generate responses by predicting the most likely sequence of words—means they can never fully distinguish between true facts and convincing fictions. They are, in essence, professional sounding boards that have no built-in commitment to reality. And when they are embedded in search engines, social media platforms, or political campaign websites, they become a primary source of information for a generation of voters who no longer read pamphlets or call election offices.
The legal framework governing Canadian elections was written for a world of paper brochures, television ads, and human canvassers. It has rules about who can say what, when they can say it, and what must be disclosed. But those rules barely touch artificial intelligence. A chatbot does not have a candidate’s name attached to it in the way a brochure does. It does not file a return with Elections Canada. It does not take responsibility for its words. When a campaign uses a chatbot to interact with voters, the lines become blurred: Is the bot a speaker? An agent? A publisher? Current law offers no clear answer. That gap means that if a voter is deceived by a chatbot—told, for example, that a candidate dropped out, or that polling stations are closed on a particular day—there may be no legal recourse. The person who programmed the bot can claim it was an automated tool with no human intent. The platform hosting the bot can claim it is merely a conduit. And the voter, left without a ballot, is told that their problem is unfortunate but not illegal. Lawmakers in Ottawa have heard this warning repeatedly from researchers, election integrity advocates, and even some technology companies. The proposal is straightforward: make AI chatbots subject to the same rules as other forms of electoral communication. That means clear disclosure when a voter is interacting with a machine. It means accountability for false information that affects voting. And it means giving election officials the power to demand corrections or removals when a bot is actively misleading the public.
But behind the policy debate is a deeply human story about trust. Democracy depends on people believing that they can participate meaningfully—that their vote counts, that the information they receive is honest, and that the system is not rigged against them. AI tools are quietly eroding that trust, not through dramatic conspiracy theories but through thousands of small, confusing moments. A student asking about mail-in ballots receives instructions that are subtly wrong, so their application is rejected. An older adult asks about transportation to the polls and is told a ride will be provided, only to wait alone at a bus stop. A new citizen asks which candidates support affordable housing and receives a summary that is entirely fabricated. Each of these people may not realize they were deceived by artificial intelligence. They are more likely to conclude that voting is too complicated, that their participation does not matter, or that the system failed them. Over time, this can create a quiet but powerful disengagement from civic life. Researchers argue that the danger is not that AI will cause a single dramatic scandal that makes headlines, but that it will produce a steady drip of misinformation that makes citizens feel helpless and cynical. The humanization of these tools—their friendly names, their conversational tone, their ability to say “I understand” at just the right moment—makes the betrayal feel personal. It is as if a trusted friend gave you directions to a place that does not exist.
That is why the call to regulate AI in elections is not merely about technical compliance. It is about protecting the basic relationship between citizens and their democracy. When people cast a ballot, they are making a gesture of faith: faith that the process is fair, that their voice will be counted, and that the information they used to make their decision was at least reasonably accurate. AI chatbots, as they currently operate, break that faith. They can impersonate human helpers, hide their limitations behind fluent prose, and shift their answers depending on subtle changes in wording—all while leaving no trail of responsibility. Bringing them under Canada’s election laws would be a meaningful step, but researchers say it is only part of the solution. Technology companies must also design these tools with built-in safeguards: refusing to answer questions about election logistics unless they can verify the information, clearly labelling AI-generated content, and directing users to official sources. Public education is equally important. Voters need to be taught that not every smiling chat window is a reliable source, and that when it comes to something as precious as a ballot, a phone call to the election office or a check of the official website is worth more than a thousand chatbot responses. This is not about being anti-technology. It is about being pro-democracy, and recognizing that every new tool brings both possibility and peril.
In the end, the debate over AI in Canadian elections is a debate about what kind of society we want to live in. Do we want a public square where machines can confidently mislead us, where the distinction between a human and a bot becomes impossible to see, and where the burden of fact-checking falls entirely on the individual? Or do we want a democracy where technology serves people without deceiving them, where rules keep pace with innovation, and where the right to vote is protected by something stronger than a good-faith hope? The answer should be obvious, but the path forward is not simple. Regulating AI will require political will, technical expertise, and ongoing vigilance. It will require recognizing that a chatbot is not a citizen, not a campaign volunteer, and not a friend—it is a tool, and like all tools, it needs boundaries. Canadian lawmakers are being urged to act now, before the next election, to ensure that AI is an aid to democracy rather than a quiet saboteur. The technology will keep evolving, and no law can predict every future development. But the core principle is timeless: when someone votes, they deserve the truth. Not an approximation of the truth. Not a convincing imitation of the truth. The truth itself. That is the promise that election law is meant to keep, and it must extend to every corner of the digital world—including the friendly little chatbot that seems so eager to help. If we lose that principle, we lose something far more important than any technological convenience. We lose the foundation of our shared civic life, and that is a cost no amount of artificial intelligence can ever repay.