This study investigates the linguistic and structural differences between human-written and AI-generated content of news articles on mental health, using The Times of India articles and AI-generated content from ChatGPT and Gemini AI. Through corpus analysis using AntConc software, the study examines keywords, collocational behaviour, sentiment, modality, and framing strategies across three distinct corpora. Using a corpus-based approach and software such as AntConc, the study examines 130 human-written articles from TOI and 130 articles written by AI between 1 January 2025 and 30 April 2025. The findings reveal that TOI articles exhibit a richer emotional vocabulary, with frequent use of terms related to personal experience and stigma, while ChatGPT and Gemini content lean towards formal, analytical language, with Gemini focusing on policy and structural framing. Sentiment analysis shows that human-written articles are more emotionally engaged, employing stronger, assertive language, whereas AI-generated articles show more neutrality and caution, particularly in terms of modality.