TRIPOD+AI and CONSORT-AI Reporting Guidelines

Reporting guidelines provide structured items authors should include so readers can understand how a study was designed and what it found. TRIPOD+AI updates reporting for clinical prediction models developed with regression or machine learning. CONSORT-AI extends CONSORT for randomized clinical trials evaluating interventions that include AI. These guides serve different study designs and do not replace one another. TRIPOD+AI supports complete reporting of prediction model development and evaluation, including participants, data sources, predictors, outcome definitions, model methods, validation, and performance. CONSORT-AI adds trial-specific details about the AI intervention, how it was used, human-AI interaction, and implementation. A checklist helps identify missing information but cannot repair a biased design, an unrepresentative sample, or an inappropriate analysis. Authors should select the guideline that fits the study and report the AI system’s intended role, version, inputs, workflow, data handling, and limitations. Editors and reviewers can use the checklists to assess transparency and reproducibility. Readers should still evaluate methods, risk of bias, and applicability to practice. Reporting quality is necessary for interpretation but is not evidence by itself that a model is safe or effective. Readers should check whether claims are consistent with the methods, whether validation data are independent, and whether limitations are reported. Transparent reporting makes it possible to identify missing details but cannot compensate for weak study design or inappropriate interpretation.