{
  "abstract": "Objectives To describe the current state of LupusGPT, a multilingual artificial intelligence (AI) tool designed to provide evidence-based information on systemic lupus erythematosus (SLE) for patients. The project aimed to use AI in patient education, challenging misinformation and enhancing equitable access to valid information on lupus.Methods Lupus Europe co-ordinated a pan-European collaboration involving patients, lupologists, and AI experts to create LupusGPT. Following EULAR 2023 recommendations on patient involvement in research, patients co-designed all stages of the tool’s development — from concept to defining ethical safeguards to testing tone and accuracy across European languages.Structured multilingual beta testing by 51 lupologists and patients evaluated accuracy, clarity, tone and medical consistency of answers across European languages. LupusGPT draws exclusively from a curated library of validated lupus resources, selected jointly by lupologists and patients, and implements a front-end interface linked to a retrieval-based data layer and an AI engine accessed via private API key, configured to minimise AI creativity and reduce risk of hallucinated content. Iterative refinements improved retrieval of specific content and allowed adjustment of potentially sensitive phrasing for patients.Results Since its launch in June 2024, LupusGPT has answered 46,217 questions across 122 countries ( figure 1).Figure 1. Global distribution of LupusGPT users (June 2024 - October 2025), visually demonstrating reach across 122 countries.Its companion feature, EasyLupus, built on the same library of valid documents, but significantly simplifying language for users with lower literacy, has answered 12,511 questions since launch in October 2024. No hallucinations (i.e. fabricated content by AI) were identified during post-launch monitoring.Abstract PO:13:338 Figure 1Conclusions LupusGPT demonstrates the feasibility of patient-led, clinician supported AI for SLE information through a freely accessible tool, allowing patients worldwide to ask questions and receive quality answers on lupus. This model is a scalable framework for digital co-production in rheumatology, while countering health misinformation on lupus and supporting equitable access to reliable, multilingual information for all patients.",
  "authors": [
    {
      "affiliations": [
        "Lupus Europe, Brussels, Belgium"
      ],
      "name": "Zoe Karakikla-Mitsakou"
    },
    {
      "affiliations": [
        "Lupus Europe, Brussels, Belgium"
      ],
      "name": "Alain Cornet"
    },
    {
      "affiliations": [
        "Lupus Europe, Brussels, Belgium"
      ],
      "name": "Jeanette Andersen"
    },
    {
      "affiliations": [
        "Centre for Musculoskeletal Research, Manchester, UK"
      ],
      "name": "Sarah Dyball"
    },
    {
      "affiliations": [
        "Centre for Musculoskeletal Research, Manchester, UK"
      ],
      "name": "Cristiana Sieiro Santos"
    },
    {
      "affiliations": [
        "UDAI – Autoimmune Disease Unit, ULS Tamega e Sousa, Penafiel, Portugal",
        "UMIB – Unit for Multidisciplinary Research in Biomedicine, School of Medicine and Biomedical Sciences (ICBAS), Porto, Portugal",
        "i3S – Instituto de Investigacao e Inovacao em Saude, Universidade do Porto, Penafiel, Portugal"
      ],
      "name": "Daniel Guimarães De Oliveira"
    },
    {
      "affiliations": [
        "Department of Rheumatology, Strasbourg University Hospitals National Reference Centre for Rare Autoimmune Diseases, Strasbourg, France"
      ],
      "name": "Laurent Arnaud"
    }
  ],
  "title": "PO:13:338 A Multilingual Artificial Intelligence Tool That Provides Quality Evidence-Based Information on Lupus for Patients (LupusGPT)",
  "uid": "3209281e-fc9e-5bf8-80f2-b5f107c7b58d"
}