PRISMA-LLM adds reporting guidance for systematic reviews that use LLMs or AI-enabled software in evidence-processing tasks. It supplements PRISMA 2020 with implementation, evaluation, limitation and reproducibility requirements.
- Reporting instructions, including the quick and expanded checklists and the requirements by method-complexity level
- Fillable reporting checklist
- Evidence and guidance informing the checklist
- Reproduce the empirical analysis
The analysis reads the harmonized corpus from the public
SciLitBench repository. One command
regenerates the paper-level analysis data, plot data, tables and figure panels;
see analysis/README.md.
framework/ Reporting instructions, fillable checklist and supporting evidence
analysis/ Reproduction code and publication-date tables
Version 0.1.0 is the initial release. Formal consensus development, usability testing, prospective validation and review by the PRISMA Executive remain future work.
Please open a GitHub issue to suggest a missing item, clarify applicability or propose a new workflow type.
The analysis code is licensed under the Apache License 2.0. The checklist, framework materials and documentation are licensed under Creative Commons Attribution 4.0 International.
If you use PRISMA-LLM, please cite:
@article{zabaleta2026prismallm,
title = {{PRISMA-LLM}: An Empirical Reporting Framework for {AI}-Assisted Systematic Reviews},
author = {Zabaleta, Miguel and Lin, Baihan},
journal = {arXiv preprint arXiv:2609.11559},
year = {2026},
doi = {10.48550/arXiv.2609.11559},
url = {https://arxiv.org/abs/2609.11559}
}