Analyzing Carbon Removal Technology Hype Cycles Through Large Language Models

This study examines whether Large Language Models (LLMs) can support the development of sentiment-based indicators for technological hype cycles, characterized by optimistic media language during hype phases and negative sentiment during periods of disillusionment. The rapid growth of digital news makes manual tracking of sentiment trends impractical. Traditional computational approaches relying on basic natural language processing often fail to capture context in long-form texts. LLMs offer a scalable alternative by enabling context-aware sentiment analysis across large collections of complex news articles. The study introduces an LLM-driven methodology to analyze temporal sentiment patterns in English-language news coverage over a 16-year period, focusing on carbon removal technologies. The approach extracts sentiments from major news sources and maps them over time to identify media-driven hype dynamics. This framework enables systematic analysis of technologies such as Bioenergy with Carbon Capture and Storage, afforestation/reforestation, Direct Air Capture, and ocean-based carbon capture. The key finding is that media sentiment tracking can complement other innovation indicators in the mapping of a Hype Cycle, revealing that CRT domain as a whole is heading towards a plateau of productivity.