Seeing is no Longer Believing: A Narrative Review of AI-Generated Images, Visual Perception, and Digital Trust
Year: 2026
Type: review
Abstract: Photographs have long served as a convenient shorthand for evidence. That habit is now under strain, because generative models can produce images that look photographic, require no camera, and can be made by anyone with an internet connection. This narrative review brings together literature published between 2016 and 2026 on AI-generated images, the way people perceive and interpret them, and what their spread means for trust in digital media. It draws on computer vision, psychology, medicine, law, philosophy, and the language-centered fields of semiotics and visual rhetoric. The reviewed studies indicate that human observers perform at or near chance when separating synthetic from authentic faces, that realistic synthetic images are often judged authentic, and that confidence in one’s own judgment is frequently misplaced. Forensic cues, including irregularities in the eyes, exist but are fragile as generators improve. Images, however, rarely travel alone: captions, narratives, source cues, and platform conventions make them multimodal texts whose meaning shapes belief as much as appearance does. The consequences extend beyond deception. Genuine images can be dismissed as fabricated, a problem known as the liar’s dividend, and fields that rely on images as data, such as ophthalmology, radiology, and scientific publishing, face new risks of falsification. Legal responses remain uneven across evidence law, cybercrime statutes, copyright, and transparency rules. The review argues that visual trust is better treated as a socio-technical judgment than as a purely perceptual ability, and that provenance standards, regulation, professional norms, critical language awareness, and visual literacy need to work together. It closes by identifying gaps in cross-cultural, multilingual, and longitudinal evidence. Keywords: AI generated images, deep fakes, visual perception, digital trust, misinformation, media literacy, image provenance
Source: Perspectives in Social Sciences
Authors: Zainab Fatima, Zaryab Khan
Institution: Government of Pakistan
Language: English
Subfield: Sociology and Political Science
Field: Social Sciences
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W7225357467