The fight against climate change demands innovative solutions, and what if many of those solutions are already hidden within the vast sea of scientific literature? Researchers are now turning to AI and large language models (LLMs) to unearth these neglected climate innovations. Imagine sifting through millions of research papers to find the needles in the haystack – the ones with the potential to revolutionize our approach to climate change. This is precisely what researchers are tackling using the power of LLMs. They've developed a system that analyzes title-abstract pairs from scientific papers, assessing their potential impact on climate change mitigation, technological readiness, and deployability. Focusing on UK-based research initially, they've built a region-agnostic workflow. They’ve trained LLMs to evaluate papers based on several key criteria, including emissions reduction potential, technology maturity, market need, and whether the research has been overlooked by traditional funding mechanisms. Surprisingly, the LLMs, especially when provided with relevant context, proved remarkably adept at identifying promising research, even outperforming human evaluators in some cases. For instance, the AI consistently identified research linked to successful climate-tech spin-out companies, often with greater certainty than human experts. However, the research also revealed that LLMs aren't perfect. They sometimes struggled to distinguish between fundamental research that could enable future technologies and research ready for immediate deployment. To refine the process, the team explored different prompting strategies and scoring systems, finding that giving the LLM more context and using a scalar scoring system (1-10 instead of yes/no) significantly improved its ability to pinpoint high-potential research. This allowed for a more nuanced ranking of research papers and helped avoid bottlenecks where numerous papers received identical scores. The initial results are promising. Applying the refined LLM approach to a larger dataset of 1000 abstracts, the AI again identified several potential game-changers, including research related to carbon capture, thermal hybrid technologies, and photovoltaics. While there's room for improvement, this research demonstrates the power of AI to accelerate the discovery of climate solutions hidden within existing scientific knowledge. Future work will focus on testing different LLMs, incorporating more sophisticated techniques like Retrieval-Augmented Generation (RAG) to give the AI access to even more context, and developing multi-agent approaches to evaluate the LLM’s outputs. They're even building a system to link research papers with the authors’ networks and social media presence to identify researchers with entrepreneurial potential. By combining the power of AI with human expertise, we can unlock the full potential of scientific research and accelerate the transition to a sustainable future.