Deepfakes in the 2025 Canadian Election: Prevalence, Partisanship, and Platform Dynamics

TL;DR

This study addresses the empirical question of how AI-generated deepfakes circulated during the 2025 Canadian federal election. Using 187,778 posts from X, Bluesky, and Reddit, it combines a high-accuracy deepfake detector with vision–language intent classification and author-leaning inference to quantify prevalence, narrative intent, and audience reach. The findings show deepfakes were present but not dominant, with platform- and leaning-specific patterns and low overall engagement, though highly realistic Fabricated content garnered relatively higher attention. The work highlights the nuanced risk where plausible, low-volume deepfakes can disproportionately influence perceptions, and it emphasizes the need for stronger, accessible detection and flagging tools to preserve trust in democratic information.

Abstract

Concerns about AI-generated political content are growing, yet there is limited empirical evidence on how deepfakes actually appear and circulate across social platforms during major events in democratic countries. In this study, we present one of the first in-depth analyses of how these realistic synthetic media shape the political landscape online, focusing specifically on the 2025 Canadian federal election. By analyzing 187,778 posts from X, Bluesky, and Reddit with a high-accuracy detection framework trained on a diverse set of modern generative models, we find that 5.86% of election-related images were deepfakes. Right-leaning accounts shared them more frequently, with 8.66% of their posted images flagged compared to 4.42% for left-leaning users, often with defamatory or conspiratorial intent. Yet, most detected deepfakes were benign or non-political, and harmful ones drew little attention, accounting for only 0.12% of all views on X. Overall, deepfakes were present in the election conversation, but their reach was modest, and realistic fabricated images, although less common, drew higher engagement, highlighting growing concerns about their potential misuse.

Figures (6)

• Figure 1: Examples of the seven intent categories used in our analysis, illustrating the range of political and non-political uses of AI-generated imagery, from defamatory and conspiratorial content to benign and artistic creations.
• Figure 2: Distribution of deepfake intents by political leaning. Bars show the prevalence of each intent category among deepfakes posted by left-, right-, and unknown-leaning accounts, revealing distinct narrative patterns—right-leaning accounts emphasize defamatory and conspiratorial content more heavily. The pie chart summarizes the overall share of deepfakes contributed by each leaning group.
• Figure 3: Prevalence of deepfake intents across platforms. Bars show the percentage of detected deepfakes assigned to each intent category on X, Bluesky, and Reddit, highlighting platform-specific differences in the types of synthetic content circulating during the election period. While this histogram is normalized per platform, the pie chart inset summarizes the overall share of deepfakes across platforms.
• Figure 4: View counts of intents. Each curve shows the ECDF: the share of posts below a given view count. Steeper early rises indicate lower exposure. Non-political deepfakes tend to reach higher view counts, while political ones remain less viewed. The legend reports the proportion of total deepfake views and the average views per post for each category.
• Figure 5: View counts by political leaning and deepfakes status. Within every political leaning, the deepfake curves rise earlier than their non-deepfake counterparts, showing that deepfake posts typically attract fewer views. The legend shows total view share and average views per post.