A month-long experiment has raised new concerns about the reliability of generative AI tools as news sources after Google’s Gemini chatbot was found to be fabricating entire news channels and publishing false reports. The findings were first reported by The Conversation, which conducted the investigation.
The experiment was led by a journalism professor specializing in computer science who tested seven generative AI systems over a four-week period. Each day, the tools were asked to list and summarize the top five news events in Quebec, rank them by importance and provide direct article links as sources. Systems tested included Google’s Gemini, OpenAI’s ChatGPT, Claude, Copilot, Grok, DeepSeek and Aria.
The most notable failure was that Gemini invented a fictitious news agency – examplefictif.ca – and falsely reported on a school bus drivers’ strike in Quebec in September 2025. In reality, the disruption was caused by the withdrawal of Lion Electric buses due to a technical problem. This was not an isolated incident. In 839 responses collected during the experiment, AI systems regularly cited imaginary sources, provided broken or incomplete URLs, or misrepresented real-world reports.
The results are important because more and more people are already using AI chatbots for messaging
According to the Reuters Institute’s Digital News Report, six percent of Canadians relied on generative AI as a news source in 2024. When these tools hallucinate facts, distort reports, or fabricate conclusions, they run the risk of spreading misinformation—especially if their answers are presented confidently and without clear disclaimers.
For users, the risks are practical and immediate. Only 37 percent of responses included a complete and legitimate source URL. While the summaries were completely accurate in less than half of the cases, many were only partially accurate or slightly misleading. In some cases, AI tools added unsupported “generative inferences,” claiming that stories “reignited debates” or “highlighted tensions” never mentioned by human sources. These additions may sound insightful, but they can create narratives that simply don’t exist.
Mistakes weren’t just limited to manufacturing
Some tools distorted true stories, such as misreporting the treatment of asylum seekers or misidentifying winners of major sporting events. Others made fundamental factual errors in survey data or personal circumstances. Overall, these issues suggest that generative AI still has difficulty distinguishing between summarizing news and inventing context.
Looking forward, the concerns raised by The Conversation are consistent with a broader industry assessment. A recent report from 22 public media organizations found that nearly half of AI-generated message responses contained significant problems, from sourcing issues to major inaccuracies. As AI tools become more integrated into search and daily information habits, the findings underscore a clear warning: When it comes to news, generative AI should at best be viewed as a starting point rather than a trusted source of information.
