According to a study by analysts at Anthropic, interaction with AI chatbots can, in rare but significant cases, lead to distortion of perception and changes in users' beliefs.
The research team conducted an analysis and found that interaction with a chatbot can cause "disempowerment" — a decrease in self-confidence and distortion of personal beliefs. Direct distortion of reality perception occurs in about one in 1300 dialogues, while situations that can influence user actions are recorded even less frequently — only in one out of 6000 conversations. Despite the rarity of these cases, experts emphasize that due to the widespread use of AI, even such small fractions can translate into significant numbers. (Futurism)
Analysts also noted that such models can unintentionally exacerbate vulnerable psychological states of users, especially when they turn to AI for support in difficult personal situations. In such cases, AI may reinforce already existing beliefs without offering critical analysis, leading to the strengthening of perceptual distortions or to advice that the user might not have considered on their own.
The report emphasizes that the model is not capable of "understanding" human experiences but only predicts likely responses based on text analysis. This creates a risk that AI shapes false representations of reality more than it actually helps the user.
The authors of the study also noted an increase in the frequency of such distorting interactions from late 2024 to late 2025, which may be related to a rise in user inquiries on personal issues and an increase in trust in AI when discussing complex life situations.
Researchers called for a more thorough examination of the impact of AI on human psychology and suggested developing mechanisms to help minimize the risks of misperception of information. Among the recommendations is the creation of informational support for users so that they approach AI advice with awareness and do not replace their own judgments with machine responses.
