Performance of ChatGPT-4o in Providing Information on Pediatric Inborn Errors of Immunity: A Cross-Sectional Evaluation

Editorial reading guide · Trust

How should I weigh a mostly good AI explanation when our child has a rare condition?

Overall quality scores were favorable, yet answers required advanced reading skills, lacked disclaimers throughout, and sometimes raised factual or safety concerns. Consistent wording and good average scores therefore did not establish dependable individual advice.

Study context

Three pediatric allergy-immunology specialists rated ChatGPT-4o answers to twenty common questions, half written from a patient/caregiver perspective, with repeated generations.

Age / stage: Pediatric inborn errors of immunity; no family participants

Read with three questions

1. Which safety concerns remained behind high overall quality ratings?
2. How did caregiver-oriented questions change comprehensibility?
3. Why is semantic reproducibility different from verified clinical correctness?

Keep the limits in view

This was one model at one time point, twenty questions, and no actual caregivers. The supplementary safety check did not verify every statement against external standards, and the rating tools were developed for static human-authored materials.