Written by Martin Ågren - May 27, 2026
Confirmation bias is:
When people tend to seek, interpret, and remember information that confirms their existing beliefs.
GenAI tools (like ChatGPT) are usually not completely objective. GenAI models tend to respond in a confirming way (rather than a questioning way).
GenAI can reinforce confirmation bias by mirroring/confirming the user’s assumptions.
This type of interaction CAN reduce the user’s critical thinking and reinforce inaccurate or incomplete beliefs.
Ask AI to challenge assumptions.
Ask for counterarguments or alternative perspectives.
Verify responses with other/independent and trustworthy sources.
Write prompts neutrally instead of using leading phrasing.
You need to be aware of the risk of reinforcing confirmation bias in the work carried out within the organization. For example, leading questions may cause AI to confirm existing perceptions instead of contributing new perspectives. This can be harmful to the organization’s necessary development and may negatively affect the quality of the information handled within the organization.
Ensure that employees have sufficient competence in understanding and preventing AI-related issues involving reinforcement of confirmation bias.
In your AI policy: Encourage the use of AI AND humans to review AI output. Ensure human review for important decisions or when important texts are produced.