During eDNAqua-Plan WP3 implementation, the use of various generative AI models (ChatGPT, Perplexity, Grok and Gemini) was evaluated to convert narrative descriptions of scientific articles into BeBOP-compatible sampling protocols (MIOP/FAIRe).
Although all models demonstrated an ability to structure information, systematic differences emerged in how they handle ambiguity, missing data, and structured metadata requirements.
The patterns observed for each model are summarised in this report.
eDNAqua-Plan
Next generation of aquatic biodiversity monitoring
Technical Resources