From Fortune · Leo S. Lo · · 3m
Four types of AI answers, only one factual
A university librarian explains how to judge AI search results by understanding what kind of answer you're actually getting.
In brief
A university librarian explains how to judge AI search results by understanding what kind of answer you're actually getting. Leo S. Lo, dean of libraries at the University of Virginia, identifies four distinct types of AI answers, factual, interpretive, constructive and strategic, each requiring different evaluation…
Google's new AI search answers
Google search now gives AI-generated answers instead of just links. When asked how much screen time is too much for teenagers, the AI cited a number but then complicated it, noting quality and balance might matter more than hours.
Medical guidance from AI
When asked should I take a daily aspirin, the AI presented medical information, warned about risks and offered tailored guidance if the user provided their age and medical history.
Different answers, different tests

Leo S. Lo, university librarian and dean of libraries at the University of Virginia, noticed these answers were different kinds of replies. Debate about AI accuracy misses that each answer type requires users to judge something different.
Four types of AI answers
Lo proposes sorting AI answers into four types: factual, interpretive, constructive and strategic. A factual claim can be checked against a source. An interpretation reflects choices about which evidence matters. A construction can be well reasoned but wrong for the person receiving it. A strategic document may not be true.
Factual answers need verification
A factual answer makes a claim that can be checked against evidence. Examples include when was the University of Virginia founded or what is the chemical symbol for gold. To determine if it is robust enough, verify the claim against an appropriate source.
Interpretive answers need context
An interpretive reply is built on evidence but has no single takeaway. Google's screen time answer said two hours was a limit, then noted the American Academy of Pediatrics says there is no exact recommended amount for teens and emphasizes the kind of screen use.
“The American Academy of Pediatrics says there is no exact recommended amount for teens and emphasizes the kind of screen use”
Constructive answers are made
Constructive answers are made rather than discovered. Ask AI to draft a cover letter, write a eulogy, suggest a lesson plan or reorganize a paragraph. There is no single correct result. Judge the response by considering purpose, audience and voice.
Strategic answers need judgment
Strategic questions ask what to do. Should I take a daily aspirin? Should I buy the house? The answers combine information with judgment about goals, risks, trade-offs and personal circumstances. Ask what the system would need to know before its advice could apply to you.
Recognize the answer type first
Before asking whether an AI answer is right, ask a more basic question: what kind of answer is this? The type will tell you what to do next. A fluent response can move between answer types without a noticeable change in voice.
“Before asking whether an AI answer is right, ask a more basic question: What kind of answer is this?”
At a glance
Leo S. Lo, dean of libraries at the University of Virginia, identifies four distinct types of AI answers, factual, interpretive, constructive and strategic, each requiring different evaluation methods.





