OTTAWA — A new study from the Angus Reid Institute reveals that while artificial intelligence is increasingly integrated into daily life, a majority of Canadians remain skeptical rather than enthusiastic about the technology.
Negative Views Dominate
According to the study released Thursday, 43% of Canadians hold negative views of AI, 24% hold critical views, 16% report positive impressions, and 17% are unsure. "One-in-three are divided between seeing the technology positively (16%) or uncertainly (17%)," the study noted, adding that this mirrors previous Angus Reid research showing Canadians hesitant to embrace AI.
The metric was derived from responses to four questions: whether AI is a force for good in society, whether everyone has equal opportunity to benefit, whether AI will create a prosperity gap, and how trustworthy respondents find AI-generated information.
Usage Rising Despite Doubts
Despite skepticism, AI use is increasing. Compared to November 2025, six percentage points more Canadians now use AI "several times a day" (10% then vs. 16% now). Daily use rose from 6% to 8%, and weekly use increased by two percentage points.
"While some Canadians are deliberately engaging with AI tools by entering prompts or queries in programs like ChatGPT or Claude, more say their exposure is passive — encountering AI-generated information or built-in AI features in apps and programs without specifically asking for them," the report stated. Fifty-two percent said their exposure is passive, 32% actively use AI-specific apps, and 15% were unsure.
Perceived Benefits and Top Concerns
Regarding benefits, 43% cited AI's ability to learn or research topics quickly, 37% said it makes information easier to access, 35% said it saves time, and 29% said it makes tasks easier. However, top concerns include people losing critical thinking skills (41%), questions about AI-generated content veracity (36%), misinformation (29%), and AI providing incorrect information (28%).
The poll was conducted May 7–11 among 1,842 Canadian adults. While margins of error cannot be applied to online panels, a comparable sample would yield a margin of ±2%, 19 times out of 20.
