A Comparative Study of Artificial Intelligence-Generated and Human Produced Content in Text, Image, and Video Domains

The proliferation of generative artificial intelligence has fundamentally altered the content creation landscape, enabling mass production of synthetic text, images, and video at unprecedented scale and fidelity. This paper presents a comprehensive, structured comparative analysis of AI-Generated Content (AIGC) and Human-Produced Content (HPC) across three primary modalities: text, image, and video. Each modality is evaluated along seven dimensions — Veracity, Engagement, Realism, Intent Coherence, Temporal/Structural Integrity, Authenticity Perception, and Societal Impact — forming the novel VERITAS evaluation framework. Key findings reveal that AI-generated text achieves near-parity in surface fluency yet lags in long-form reasoning and factual grounding, with hallucination rates ranging from 0.7% to over 10% across leading models. AI-generated images, driven by diffusion architectures such as FLUX.1 and Stable Diffusion 3.5, achieve photorealism sufficient to reduce human detection accuracy to 63.7%. AI-generated video, exemplified by Google Veo 3 and OpenAI Sora, struggles with physical plausibility and long-range temporal coherence despite achieving short-form realism. Deepfake videos grew 900% between 2023 and 2025, while consumer preference for human-created content rebounded to 74% in 2026. We introduce the TRACE detection taxonomy, propose policy and technical recommendations, and identify open research challenges in cross-modal forensics, adversarial robustness, and the preservation of human creative value.