Artificial intelligence tools such as ChatGPT, Claude, and Gemini are becoming increasingly popular sources of information.
Many people now use these systems not only to answer questions but also to check facts and stay informed about current events.
Surveys show that about one in five American teenagers regularly use AI chatbots to get news, while one in four young adults have tried using them for that purpose.
However, a new study from the MIT Media Lab suggests there may be an unexpected downside to relying too heavily on AI for fact-checking.
The researchers found that while AI can help people identify misinformation in the short term, it may reduce their ability to spot false information on their own over time.
The study followed 67 participants over four weeks. During the experiment, people were asked to judge whether news headlines paired with images were real or fake.
When participants received help from an AI chatbot, their accuracy improved by about 21%. This finding supports previous research showing that AI can be useful for evaluating information and reducing belief in false claims.
But the researchers discovered something surprising when the AI assistance was removed.
By the end of the four-week study, participants were significantly worse at detecting misinformation without help. Their performance dropped by 15 percentage points compared with their ability at the beginning of the study.
Even more striking, some participants believed they were becoming better at spotting fake news when the opposite was happening. About one-quarter of participants reported increased confidence in their skills despite their declining performance.
Researchers describe this effect as part of the “AI dependency paradox.” The idea is that while AI can make tasks easier and improve performance in the moment, relying on it too much may weaken people’s own abilities. Similar patterns have been seen with other technologies. Calculators can reduce mental arithmetic skills, and GPS navigation can make people less reliant on their natural sense of direction.
The researchers found that some participants gradually stopped thinking critically and instead accepted the chatbot’s suggestions without much independent analysis. They labeled these individuals “Dependency Developers.”
One participant noted that the chatbot encouraged checking multiple sources but did not teach them how to carefully examine the images themselves or understand the broader context of a story.
The researchers also warned that AI systems can struggle during fast-moving, emotionally charged news events. During major breaking news stories, such as political violence or international conflicts, chatbots may spread inaccurate information because they rely on patterns in existing data rather than true understanding.
The team believes the key is not to eliminate AI use but to design AI tools differently.
According to the study, AI systems that simply provide answers are more likely to create dependency. In contrast, systems that ask questions and guide users through the reasoning process can help people develop stronger critical-thinking skills. This teaching approach, often called the Socratic method, encourages users to think through problems themselves rather than relying on instant answers.
The researchers say society needs a new form of AI literacy. As AI becomes a larger part of everyday life, people must learn when to use these tools and when to rely on their own judgment.
The ability to question information, think critically, and form independent opinions remains an essential skill—one that no chatbot should completely replace.
Source: MIT.
