CARE is an approach developed by Leo S. Lo for critically examining AI-generated answers. Introduced in a conceptual commentary in The Journal of Academic Librarianship in 2026, it connects an answer’s claims, evidence, assumptions, and usefulness to the reader’s purpose.
How to use CARE to evaluate an AI answer.
An AI answer may look complete before you have examined its evidence, assumptions, or fit with your situation. CARE gives you four moves for working with that answer.
01 Classify
Notice the kind of answer you are reading. Is it giving facts, interpreting something, creating a draft, or proposing a course of action?
02 Assess
Ask what is present, what is missing, and what assumptions shape the answer. Identify the part that matters for your next decision.
03 Review
Reconnect the answer to evidence. Check one or two central claims against relevant sources. Note what the source establishes and what remains unresolved.
04 Enhance
Revise, extend, or qualify the answer using what you learned. Make the result fit your purpose, evidence, and context.
Answer Typography: four broad answer patterns.
- Factual
- Statements, definitions, dates, names, or numbers that can be checked.
- Interpretive
- Explanations and comparisons that select a perspective or account of why something happens.
- Constructive
- Drafts, outlines, and other content intended to help you produce something.
- Strategic
- Plans and recommendations about how to act or proceed.
These are broad patterns, called Answer Typography in the paper. An answer may mix them.
Illustrative example
Examine a claim about peer review.
Suppose an AI answer says, “Peer-reviewed articles are always reliable.”
- Classify: This is a factual claim, with an absolute word: “always.”
- Assess: It treats a review process as a guarantee. It leaves out the quality of the methods, later corrections, and disagreement.
- Review: Open the research being discussed. Examine its methods and limitations, check for corrections or retractions, and compare relevant evidence.
- Enhance: Replace the blanket assurance with a qualified explanation of what peer review contributes and what still needs checking.
This example illustrates the approach. It does not report a study or a measured learning outcome.
How CARE differs from prompting guidance.
CLEAR helps you formulate and revise a request. CARE helps you examine the answer that comes back. The two can be used together as you revise both your question and your understanding.
What CARE does not establish.
CARE offers a conceptual approach to critical engagement with AI answers. It does not certify an answer’s accuracy, replace subject expertise, or by itself demonstrate improved learning outcomes.
Original publication and citation
Lo, L. S. (2026). The CARE approach for academic librarians: From search first to answer first with generative AI. The Journal of Academic Librarianship, 52(1), Article 103186. https://doi.org/10.1016/j.acalib.2025.103186.
Read the original source