University of Virginia Dean Proposes AI Answer Typography

University of Virginia Dean of Libraries Leo S. Lo proposed an answer typography designed to help users evaluate AI-generated responses, such as those found in Google search results. Lo argues that because AI delivers information in a consistently fluent and authoritative voice, users often struggle to determine the appropriate method of verification for different types of content.

The framework categorizes AI answers into four distinct types. Factual answers make claims that can be verified against evidence. Interpretive answers are based on evidence but lack a single takeaway, requiring users to consider omitted evidence or alternative interpretations. Constructive answers, such as drafted letters, are created rather than discovered and should be judged by their purpose and intended audience. Strategic answers provide guidance based on goals and risks, requiring users to assess their personal context and consult qualified professionals.

Lo emphasizes that accuracy is only one metric for trust. By first identifying which of the four categories a response falls into, users can apply the correct judgment criteria—whether that involves checking primary sources, analyzing emphasized evidence, or seeking professional advice.