인공지능 생성 뉴스 제목과 사람이 만든 헤드라인의 감성, 언어전략, 자극성 비교 분석

초록

This study aims to compare news headlines written by generative AI (ChatGPT) with those crafted by human journalists, exploring the potential and limitations of AI-based news automation. A total of 48 real news articles published in June 2025 were collected. For each article, both the original human-written headline and an AI-generated headline (using ChatGPT with a fixed prompt) were analyzed. The comparison focused on five dimensions: sentiment score (via KoBERT), presence of sensational words, sentence length, part-of-speech composition, and sentence-ending types. The results reveal that AI-generated headlines tend to maintain a neutral and explanatory tone, while human-written headlines employ more emotional language and strategic expressions to draw reader attention. Qualitative comparisons further show that humans more actively use symbolic, emotional, or rhetorical expressions to elicit empathy and engagement. This study contributes to understanding how AI differs from human judgment in headline composition and suggests practical implications for ethical AI use and content design in journalism.

키워드