Artificial Intelligence: How do we know that a photo is AI?

An online dispute originated after the post of the deputy spokesperson of ELAS George Balatsoukas on Facebook, which included an AI-generated photo of Prime Minister Kyriakos Mitsotakis hugging PSD leader Niko Androulakis and opposition leader Olga Markogiannakis. A response was provided by the PSD Spokesperson Olga Markogiannakis, who countered the post by suggesting the image was fake news and pointing out an alternative photo involving Mr. Tsipras and Panos Kammenos. The discussion led to a reflection on the risks posed by Artificial Intelligence in online content, emphasizing that easily accessible AI-generated media can be indistinguishable from reality. The text outlines several visual and auditory cues—such as distortions, unnatural skin texture, metadata, and inconsistencies in shadows and perspective—that can indicate artificial creation. The source also details seven specific signs to help readers assess the authenticity of images and media encountered online.

The scenario illustrates the friction between easily disseminated, emotionally resonant visual claims and the increasing technological capacity for synthetic media. The conflict pivots not on the original political assertions, but on the epistemic crisis induced by AI-generated imagery—a shift where verifiable reality becomes an elusive object requiring active verification rather than passive acceptance. The process of identifying AI artifacts, as detailed by the GIJN guide, reveals that detection is less about spotting obvious errors and more about scrutinizing systemic inconsistencies in how images are constructed: light physics, geometric alignment, temporal consistency, and behavioral realism. This dynamic shifts public discourse from accepting imagery at face value to engaging in a constant, skeptical audit of digital evidence. The underlying pattern suggests a struggle where the speed of content creation outpaces the societal capacity for critical verification, creating an environment where authenticity is weaponized through plausible deniability. The question becomes whether reliance on external markers (like metadata or AI detection) constitutes a new form of authority in establishing truth when visual and auditory evidence can be manufactured so seamlessly. What structures are needed to maintain shared epistemological ground when the very foundations of visual proof are negotiable?