Visualizing the Adoption of Large Language Models across Sociology Subfields

Generative large language models (LLMs) entered scientific writing at scale after the public release of ChatGPT in November 2022, but their adoption in sociological research remains largely unmapped. This visualization estimates the share of LLM-modified sentences in 1.49 million abstracts from sociology papers indexed in OpenAlex between January 2018 and December 2025, using the distributional alpha estimator developed by Liang and colleagues. By tracing change over time, across substantive subfields, and across author groups, it provides a field-level view of how rapidly LLM-assisted writing has diffused through sociology. The visualization highlights three patterns. First, estimated LLM modification rose from near zero before late 2022 to 24.3 percent of abstract sentences in July–December 2025. Second, adoption varies substantially and persistently across subfields: empirical and applied areas show the highest estimated uptake, while theory-oriented subfields remain lowest. Third, data from the post-ChatGPT era show that LLM-assisted writing is also more prevalent among younger scholars, authors from less prestigious institutions, and publications in less prestigious outlets. Together, these patterns show that LLM use in sociology has become widespread but remains unevenly distributed across scholarly communities.