A systematic literature review of LLM-based data visualization

Abstract

The increasing popularity of Large Language Models (LLMs) has sparked growing interest in their application across a wide range of tasks, including data visualization. This study presents a Systematic Literature Review (SLR) aimed at investigating how LLMs are exploited to create data visualizations, addressing the research question: How can LLMs be used to create data visualizations? To this end, we identified ten dimensions of analysis, covering the study context, visualization-related aspects, and LLM-related characteristics. The visualization dimensions examine the types of visualizations produced, their outputs and interactivity, the underlying visualization theories or grammars, the intended target users, and the evaluation methodologies adopted. The remaining dimensions focus on the role and configuration of LLMs within the proposed approaches. Our findings provide a structured overview of current research trends, highlight how LLMs are currently leveraged to support or automate visualization creation, and identify open challenges and research gaps, offering directions for future work at the intersection of generative AI and data visualization.