Effective communication is a key enabler of climate change adaptation in agricultural systems. However, different actors frame adaptation, transition, and transformation in varied ways, influencing how change is understood and acted upon. This study uses Natural Language Processing (NLP) and machine learning to analyse how climate adaptation is communicated across five actor groups in Aotearoa New Zealand: media, farm advisory services, researchers, Indigenous Māori, and government. We apply topic modelling, sentiment analysis, collocation network analysis, and word embedding models to five purpose-built corpora to identify dominant themes, emotional tones, and framings of responsibility and agency. This methodological approach enables systematic, large-scale comparison of discourses, offering insights into how adaptation narratives evolve and diverge across sectors. Our findings highlight both overlaps and tensions in how different actors communicate about climate risks and responses. For example, while some narratives emphasise innovation and opportunity, others centre on uncertainty or systems-level transformation. These differences have practical implications for how messages are received, interpreted, and acted upon by farmers and stakeholders. By identifying areas of alignment and dissonance, we show how NLP tools can support the design of more targeted and effective communication strategies. This contributes to methodological innovation in climate communication research and offers practical value for policymakers, advisors, and communicators seeking to accelerate adaptation through more resonant messaging. Our study demonstrates the potential of data-driven discourse analysis to support climate-resilient agricultural futures.