Generative AI and the Reconstruction of Visual Reality: Creative Intervention and the Boundaries of Authenticity in Documentary Media

Abstract

Generative artificial intelligence (GenAI) is fundamentally changing how documentary images relate to the reality they represent. Documentary has never been a simple reproduction of reality; filming, editing, and narrative construction have always involved subjective creative choices. However, GenAI introduces a further paradigm shift by allowing audiovisual material to be generated or reconstructed without a corresponding act of photographic recording. This technological leap makes the traditional relationship between photorealism and indexicality significantly less stable. Consequently, an image may look convincingly real without being a photographic record of an actual event. Furthermore, GenAI radically alters creative practice. Prompting, selecting, correcting, regenerating, and sequencing generated material can distribute creative agency across human creators, generative systems, training data, and complex production processes. Yet, distributed creative agency does not mean that factual responsibility can be distributed in the same manner. This article therefore considers documentary authenticity through four related dimensions: referential authenticity, procedural authenticity, narrative authenticity, and factual responsibility. By examining contemporary case studies and theoretical frameworks in media ethics, this research highlights the evolving challenges in visual documentation. Rather than making documentary truth irrelevant, GenAI makes it more necessary than ever to explain how an image relates to historical reality, how it was produced, how it functions within a documentary narrative, and who is ultimately responsible for its factual claims.