Nova raziskava: Googlov pregled umetne inteligence krepi nekatere teorije zarote

New research suggests that Google’s AI Overviews, which summarize search results, may unintentionally reinforce certain conspiracy theories. According to a study published in Media International Australia, these AI-generated summaries can present misleading or incomplete information, particularly when users search for terms associated with conspiracy theories. The findings raise concerns about how search algorithms might influence public perception and the spread of misinformation. The study examined two specific conspiracy theories: “chemtrails” and “15-minute cities.” Chemtrails, a belief that aircraft contrails are part of a government-led weather manipulation program, have circulated since the 1990s. Researchers previously criticized Google for promoting such ideas, prompting the company to adjust its search results to reduce exposure to harmful misinformation. However, the latest research indicates that similar issues persist with newer, more recently emerged conspiracy theories. “15-minute cities,” originally a concept aimed at creating walkable urban environments, have been co-opted by some conspirators who claim the plan represents a globalist effort to control populations through surveillance and technology. The researchers tested how Google’s AI Overviews respond to searches involving these conspiratorial narratives. They created search queries that reflected both mainstream and conspiratorial viewpoints and monitored the resulting AI summaries over a period of a week in early 2026. The study found that search queries containing pro-conspiracy language generated significantly different AI summaries than neutral or standard searches. For instance, queries like “15 minute city segregation” produced responses that subtly reinforced the idea that urban planning initiatives were part of a larger, secretive agenda. In contrast, searches for non-conspiratorial versions of the same topics generally received more balanced or debunking responses. In the case of chemtrails, the AI Overviews were less likely to directly promote the conspiracy theory. Instead, they often mentioned the theory briefly before redirecting users to sources that challenge its validity. However, the researchers noted that these debunkings were typically superficial, offering little in the way of detailed evidence or reasoning to counter the conspiracy. The situation was more concerning for the “15-minute cities” narrative. Some AI Overviews actively presented the false claims as part of a legitimate debate, rather than clearly refuting them. One example involved a query about the relationship between 15-minute cities and smart cities, where the AI summary described the two concepts as complementary but emphasized the control aspects that critics highlight. This approach could mislead users into thinking there is a valid argument supporting the conspiracy. The researchers argue that current safeguards are insufficient to prevent AI Overviews from reinforcing harmful narratives. They emphasize the need for stronger mechanisms to ensure that search results provide accurate, comprehensive, and logically sound information, especially when dealing with topics that are prone to misinformation. As the use of AI-driven search summaries becomes more widespread, the implications of this research grow more urgent. Users rely heavily on these summaries for quick access to information, yet the study reveals that the format itself can shape understanding in unintended ways. The findings underscore the importance of ongoing scrutiny and improvements in how search engines handle complex, potentially misleading topics.