Custom GPTs to aid in compliance checking for reporting standards in academic publishing

Area of Science:

• Biomedical research
• Scientific reporting standards
• Artificial intelligence in science

Background:

• Despite numerous reporting standards, incomplete methods reporting is common due to compliance challenges.
• Existing checklists are tedious, inconsistently enforced, and poorly integrated into scientific writing workflows.
• Better reporting is essential infrastructure for reproducible biomedical research.

Purpose of the Study:

• To explore a pragmatic approach for improving compliance with reporting standards.
• To develop interactive, AI-driven guidance tools for scientific manuscript preparation.
• To translate static reporting checklists into dynamic, integrated support systems.

Main Methods:

• Development of three custom GPT-based AI assistants (compliance copilots).
• Copilots are aligned with specific reporting frameworks: Good In Vitro Reporting Standards (GIVReSt), ToxRTool, and ARRIVE.
• Tools provide real-time feedback, flag missing information, and offer suggestions during manuscript writing.

Main Results:

• Early benchmarking indicates AI copilots can approximate expert judgments for reporting checklist items.
• The AI approach achieves high consistency and significantly reduces the time required for compliance checking.
• Analysis highlights the importance of "strict" vs. "lenient" interpretations and the role of AI as decision-support.

Conclusions:

• AI can transform reporting standards from aspirational goals into routine practice.
• Making rigorous reporting the path of least resistance is key to improving scientific transparency.
• The cultural integration of AI is more critical than the technology itself for enhancing reporting standards.