What the study did
Researchers compared how five generative AI tools (ChatGPT, Gemini, Copilot, Meta AI, and Perplexity) draft informed consent documents across six mock research scenarios, each document comprising a participant information sheet and a consent form. Three internal assessors, two of them blinded, scored every document against a rubric validated by three ethicists and aligned with ICMR, Declaration of Helsinki, and ICH-GCP standards.
To analyze the text itself, Evidano was used for content analysis. It surfaced recurring omissions and risky phrasing across all five tools, including placeholder IRB contacts, missing compensation clauses, and blanket confidentiality assurances, which the authors tabulated with representative excerpts.
Published open access in Perspectives in Clinical Research, Vol. 17(2), 2026, the official journal of the Indian Society for Clinical Research, with AI tool use disclosed in the methods and institutional ethics committee clearance on record.
How Evidano fit the workflow
- 1Collect the documents
30 AI-drafted consent documents across six research scenarios.
- 2Run content analysis in Evidano
Evidano analyzed the documents for omissions, risky phrasing, and language patterns.
- 3Surface the findings
Recurring omissions and risky phrasing were identified, with representative document excerpts.
- 4Human review and reporting
Three trained assessors scored the documents; results were reported in the published study.
“content analysis was performed using the AILYZE tool to identify omissions, risky phrasing, and tool-specific language patterns”
“Content analysis from the AILYSE tool revealed recurring omissions and risky phrasing across all tools”
AI tool use was disclosed in the methods and cited as a reference in the paper.
Evidano supported the analysis; trained human assessors scored every document.
Content analysis of drafted documents, policies, and reports uses the same workflow as interview data.