TITLE: Study: AI Chatbots Sway Canadian Voters More Than Americans

AI Chatbots Demonstrate Strong Persuasive Power on Voters

A recent study has found that interactions with AI chatbots can significantly alter voting intentions, with Canadians proving more susceptible to influence than their American counterparts. Published in the journal Nature, the research involved 1,530 Canadian participants and examined the role of generative AI in political persuasion during the April federal election.

The experiment exposed participants to chatbot conversations tailored to key policy issues they identified as priorities among 17 options. All interactions occurred in English, with no geographic breakdown provided for participants. Researchers observed that brief engagements—lasting six to eight minutes—were sufficient to shift voter preferences.

In the Canadian segment, conducted in the election's final week, the chatbots achieved notable success. Under a pro-Carney condition, one in nine respondents changed their voting intention. In the pro-Poilievre scenario, one in 13 switched. These rates suggest that targeted AI deployment could sway close races, potentially flipping outcomes in specific ridings.

Comparative Impact: Canada vs. United States

The study also included a U.S. experiment from fall 2024, where one in 21 participants switched to supporting Kamala Harris after chatbot interaction, and one in 35 shifted toward Donald Trump. This indicates a lower persuasion rate in the U.S., where the effect was almost three times smaller than in Canada.

One explanation for the disparity lies in the political environment. Americans face near-constant election-related content, potentially desensitizing them to persuasion attempts like ads. In contrast, Canada's less saturated campaign cycle may leave voters more open to new influences, such as adaptive AI dialogues.

The chatbots' effectiveness stemmed from their ability to customize arguments based on individual responses. When restricted from citing facts and evidence, the persuasion rate dropped by more than half, underscoring the importance of data-driven arguments in AI's influence.

Mechanisms of Persuasion and Broader Implications

Unlike static political ads, AI chatbots engage users in dynamic, personalized conversations. This interactivity allows for real-time adaptation, making the experience more compelling than traditional media. The study concludes that such technology 'can meaningfully impact voter attitudes,' with authors warning of its likely role in future elections and profound democratic consequences.

While the Canadian experiment involved a broad national sample, researchers doubt it could have altered the actual election results, given the scale. However, in tight ridings decided by narrow margins, even small shifts could prove decisive.

Regulatory gaps exacerbate these risks. Canada imposes strict rules on election advertising but lacks specific guidelines for AI use. Elections Canada notes that AI could violate laws if used to impersonate officials or disseminate false materials from parties or candidates. No evidence suggests AI affected the last election's outcome, based on June complaints reviewed by the Office of the Commissioner of Canada Elections.

Chief Electoral Officer Stéphane Perrault has proposed amendments to election laws, including requirements for transparency markers on AI-generated communications and directives for chatbots to link to official information sources.

Expert Perspectives and Global Context

Fenwick McKelvey, an associate professor at Concordia University specializing in communication studies and co-director of its applied AI institute, commended the research for quantifying AI's persuasive potential. He noted that parties in countries like Mexico already deploy chatbots for voter outreach.

A major concern is the integration of AI with voter databases held by political parties, which are exempt from privacy laws. McKelvey advocates extending privacy protections to these databases and mitigating AI harms in advertising to prevent unconsented data uses.

The study arrives amid growing scrutiny of AI in elections. In the Netherlands, a watchdog has deemed AI chatbots unreliable for voting advice. Canada's CRTC recently updated Canadian content definitions to limit AI involvement, reflecting broader efforts to balance innovation with safeguards.

As AI tools evolve, their deployment by campaigns could reshape electoral dynamics. The research emphasizes the need for proactive measures to preserve democratic integrity, ensuring that technological persuasion does not undermine voter autonomy.