Artificial intelligence (AI) is rapidly transforming medicine, and a recent review in The Journal of Allergy and Clinical Immunology (February 2026) explores it.
Recent AI Breakthroughs and Their Relevance
Advances in large language models (LLMs) (e.g., ChatGPT-like systems), multimodal AI (handling text, images, and other data), and AI agents (autonomous systems for decision-making and task execution) are driving progress.
These tools excel at processing unstructured data like clinical notes, predicting disease progression, matching patients to trials, and even analyzing genetic or proteomic "text."
In allergy/immunology, early applications include:
- Predicting asthma from electronic health records
- Diagnosing allergies via epigenetic patterns
- Detecting eczema or atopic dermatitis from skin images
- Interpreting skin prick tests, conjunctival provocation photos, or airway CT scans with high accuracy
- Monitoring airborne allergens (e.g., pollen) using low-cost sensors
- Enhancing patient education, health literacy (e.g., explaining asthma info), and administrative tasks like digital scribing
Despite over 1,000 FDA-approved AI medical devices by mid-2024, none specifically target allergy and immunology.
Key Challenges to Implementation
Data issues: Privacy risks, bias, incomplete records, and lack of standardized/large datasets.
Technical limits: Hallucinations (confident but wrong outputs), domain-specific knowledge gaps, and performance drift over time.
Practical barriers: Workflow integration, clinician trust/training needs, regulatory evolution (e.g., EU AI Act), liability concerns, and unclear reimbursement.
The "AI chasm": Most work stays in research/proof-of-concept; few randomized trials demonstrate real clinical value.
AI holds potential to improve diagnosis, personalize treatment, reduce administrative burden, enhance prevention (e.g., pollen forecasting), and accelerate research in allergy and immunology.
The full open-access review is available in JACI (DOI: 10.1016/j.jaci.2025.08.022).
https://www.jacionline.org/article/S0091-6749(25)00939-X/fulltext