New AI pipeline reveals climate discourse split between Meta ads and Bluesky posts
Researchers developed an LLM-powered tool to compare paid vs. organic climate conversations across platforms.
A comparative analysis of climate discourse across paid ads on Meta and public posts on Bluesky from July 2024 to September 2025 uses an interpretable thematic discovery pipeline that clusters texts by semantic similarity and labels them with LLM-generated themes. Paid ads focus on strategic promotion of solutions in a formal, forward-looking register, while organic posts center on systemic critique in a crisis-oriented, scientifically grounded one. The themes are also evaluated for utility in downstream stance prediction and theme-guided retrieval tasks.
- LLM-powered pipeline clusters climate texts by semantic similarity and labels themes without requiring any predefined topic seeds.
- Meta ads promote specific solutions in a formal tone; Bluesky posts critique systems in crisis-oriented, scientific language.
- Discovered themes improved downstream stance prediction and theme-guided retrieval across both platforms.
Why It Matters
Professionals in policy, marketing, and social science gain a scalable tool to decode cross-platform narrative manipulation.