Impersonal Statements: LLM-Era College Admissions Essays Exhibit Deep Homogenization Despite Lexical Diversity

Concerns that AI will homogenize human thinking—and homogenization findings in lab contexts—appear at odds with findings that AI-assisted content is judged as highly original. Using a multi-method approach that combined a controlled experiment with two large-scale natural experiments in over 160,000 U.S. college admissions essays, we observed a counterintuitive pattern: AI-assisted essays and essays written after the release of ChatGPT showed greater lexical diversity, but the increasingly diverse words actually expressed increasingly homogeneous ideas. This “paradoxical homogenization” pattern was robust across demographic subgroups but emerged disproportionately among applicants outside racial and linguistic mainstreams. Event-time analyses and robustness checks ruled out pre-existing trends and pandemic-related confounds. The findings suggest that surface-level diversity of AI-generated content masks deeper constraints on the breadth of human ideation. Extending lab-based AI homogenization research, paradoxical homogenization appears to have reached real-world, high-stakes contexts where writing is expressly personal and AI use is explicitly prohibited.