A new study from the MIT Media Lab has raised concerns about the long-term impact of relying on artificial intelligence tools for fact-checking and news verification, finding that frequent use of AI assistants may reduce users’ ability to independently identify misinformation.

The open-access research tracked 67 participants over four weeks as they evaluated news headline-image pairs with the assistance of large language models (LLMs) such as ChatGPT, Claude and Gemini. While AI support initially improved performance, researchers found that participants became less effective at spotting false information when the AI assistance was removed.

According to the study, participants were 21 per cent more accurate at identifying fake news while using AI chatbots. However, by the end of the month-long experiment, their unaided performance on new news items had declined by 15 percentage points compared with their baseline scores.

Researchers describe the phenomenon as the “AI dependency paradox,” where tools designed to improve decision-making can gradually erode the very skills they are meant to support. The findings mirror broader concerns around technological deskilling, similar to the way calculators have reduced mental arithmetic abilities or GPS systems have affected navigational skills.

The study identified a subgroup of users termed “Dependency Developers,” who increasingly shifted from active evaluation to passive acceptance of AI-generated guidance. Notably, some participants believed their misinformation-detection skills had improved even as their actual performance deteriorated.

The research team argues that the design of AI systems plays a crucial role in determining whether they act as learning aids or cognitive crutches. Systems that provide direct answers were found to encourage dependence, while those employing Socratic questioning and guided prompts helped users develop stronger independent reasoning skills.

Presented at the 2026 CHI Conference on Human Factors in Computing Systems, the study underscores the need for educators, technology developers and policymakers to carefully consider how AI tools are integrated into learning and information-verification workflows.

The researchers emphasize that while AI can be a powerful aid in combating misinformation, preserving critical thinking and independent analysis remains essential in an increasingly AI-mediated information ecosystem.