https://link.springer.com/article/10.1007/s11186-026-09690-2
I cannot say much about the text analysis that is LLM-based, but some thoughts on other elements of the analysis:
a) Analyzing ideological stance of abstracts from 1960-2024 against a US 2025 ideological spectrum 1/

The ideological orientation of academic social science research 1960–2024 - Theory and Society
This study analyzes approximately 600,000 English-language social science abstracts published between 1960 and 2024 to estimate the long-run ideological orientation of disciplinary research output. Large language models (LLMs) were applied to each abstract using a fixed 2025 U.S. ideological spectrum, enabling consistent coding across six decades. Five key findings emerged. First, roughly 90 percent of politically relevant social science articles leaned left 1960–2024, and the mean political stance of every social science discipline was left-of-center every year during the period. Second, all disciplines showed leftward movement between 1990 and 2024. Third, policy-proximal disciplines generally showed limited rightward moderation between roughly 1970 and 1990, though policy-distal disciplines did not. Fourth, disciplines with greater leftward orientation generally displayed greater ideological homogeneity Fifth, sociocultural content was more consistently left-leaning than economic content, and that gap widened over time. Robustness checks using a wide assortment of alternative datasets and analytical methodologies indicated that these findings were unlikely to be artifacts of idiosyncratic assumptions. Methodologically, the study demonstrates the capacity of LLM-based text classification to deliver reliable, large-scale ideological measurement over time, a task previously impractical with human coding alone. Taken together, the analysis provides the first systematic, cross-disciplinary evidence of the long-run political orientation of anglophone social science scholarship, revealing both the persistence and the intensification of its leftward tendencies, particularly in sociocultural domains.