Generative artificial intelligence for climate information services

The escalating global climate crisis, marked by rising temperatures and extreme weather events, demands a fundamental shift in how Climate Information Services (CIS) are delivered to enhance climate resilience. Around the world, many local communities lack access to relevant climate information due to ineffective communication and integration channels. Most research and efforts related to CIS focus on enhancing scientific prediction and observation systems, rather than making these systems user-friendly by involving communities as end users. Generative Artificial Intelligence (GenAI), especially its large language models (LLMs), has emerged as a transformative technology with promising applications in CIS. While individual studies show potential, their findings are fragmented. We conduct a systematic literature review (2022–2025) following PRISMA guidelines to identify relevant studies on GenAI, particularly LLMs such as ChatGPT, in climate-related contexts. The chosen timeframe is based on ChatGPT’s 2022 release. After a thorough search, we found 281 articles; after screening, 19 were chosen for detailed analysis. The results indicate that implementing GenAI in CIS is vital for improving early warning systems, fostering community engagement, generating climate data, enhancing weather forecasts, and mapping local climate vulnerabilities. However, few studies examine user experience, which is essential for building trust and encouraging widespread adoption. Therefore, inclusive, bottom-up approaches should be adopted to promote community participation in deploying GenAI technologies for CIS in local settings. Despite its promising potential, GenAI also poses challenges, including a lack of transparency (black-box systems), misinformation, and digital accessibility issues. Therefore, the initiation and deployment of GenAI models should also consider their associated challenges.