Assessing ChatGPT for taxonomic and floristic studies

The advancement of biological sciences has long been linked to technological progress. ChatGPT, a generative artificial intelligence chatbot, produces human-like conversational responses and offers potential support for research across diverse scientific disciplines. However, its utility in natural sciences, particularly botanical research, remains largely unexplored. This study systematically evaluates ChatGPT’s performance in plant identification, creation of taxonomic keys, morphological description analysis, distribution mapping, and image-based species recognition. Across multiple tests involving taxa from different regions, the chatbot frequently produced inconsistent or incorrect outputs, including misidentification of species, erroneous synonymy assignments, fabricated references, and non-functional visualisations. Performance was especially limited for closely related species, hybrids, and microphotographs, while genus-level identification and text refinement were comparatively..., , # Data from: Assessing ChatGPT for taxonomic and floristic studies

Dataset DOI: [10.5061/dryad.6t1g1jxdf](https://doi.org/10.5061/dryad.6t1g1jxdf)

## Description of the data and file structure

These are supplementary materials to the article \"Assessing ChatGPT for taxonomic and floristic studies\" in the Nordic Journal of Botany.

## Code/software

Excel, PDF, R

## Access information

Other publicly accessible locations of the data:

* [https://github.com/Olene/ChatGPT-in-botanical-researches](https://github.com/Olene/ChatGPT-in-botanical-researches)

Data was derived from the following sources:

* As chats of the ChatGPT to test AI for botanical research (30 questions were asked in 8 countries) and their analysis.

## **Files**

**Script_Similarity_Matrix.txt (694 B)**

This R script for difining of text similarity (%) of analysed ChatGPT chats using TF-IDF (Term Frequency – Inverse Document Frequency). 

**Supp._Mat._1-7.zip (67.79 MB)**

This .zip folders contains 7 PDF files with..., Some files may contain the names of co-authors of the dataset.,