Large Language Models as Consumer Lab-Report Interpreters: A Design Case Study of an AI Tool for Mold and Biotoxin Testing
Bisht, H. · 2026 · Zenodo · Journal article · CC BY 4.0
People increasingly read their lab results before a clinician does, and many now hand them to a general-purpose model to make sense of. This paper looks at the narrower case of an interpreter built for one domain, using our own free mold and biotoxin report tool as the study subject. It examines the design decisions that carry the risk: narrowing scope, framing output as educational rather than diagnostic, storing nothing, guiding redaction, and defending against prompt injection in uploaded documents. It sets those against the literature on model accuracy and hallucination in clinical interpretation, and against the regulatory line between clinical decision support and a regulated medical device.
DOI: 10.5281/zenodo.21851379