IPHS 300: AI for Humanity
Document Type
Poster
Publication Date
Spring 2025
Abstract
With nearly 30% of people ages 14-22 reporting frequent AI use, increasing AI assistance in college admissions essays poses a new consideration with ethical AI use (Rubin et al., 2024). This study investigates how well three generative AI models write college admissions essays: ChatGPT o3, Gemini 2.5 Pro, and Grok. Each model responded to a Common App 2025-2026 application prompt as three distinct applicants: Elijah, Grace, and Malik (Common App, 2025). Each student represented a common college applicant archetype: Elijah, an academically successful dual sports recruit; Grace, a near-perfect award-winning scholar; and Malik, a gifted student from an underrepresented background. Claude’s 3.7 Sonnet was modeled as an experienced admissions officer and followed a strict, eight-part grading rubric. Results showed apparent performance differences across models and character types, with ChatGPT consistently outperforming the other models. Malik’s profile produced the strongest essays across all models, while Elijah’s essay showed consistent weaknesses in grading criteria. These findings raise important questions about AI’s ability to replicate authenticity and its implications for admissions equity.
Recommended Citation
Mangine, Adrian, "Authenticity Under Review: How Well Does GenAI Write College Admissions Essays?" (2025). IPHS 300: AI for Humanity. Paper 59.
https://digital.kenyon.edu/dh_iphs_ai/59
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.