Synthetic Media and the Epistemics of Public Communication: A Theoretical Account of Deepfakes, Authenticity Infrastructure and Collective Belief

For most of the twentieth century, audio and video recordings were among the few kinds of media that publics accepted as evidence without asking who stood behind them. Generative models have ended that arrangement. Convincing synthetic voices, faces and scenes can now be produced cheaply and at scale, and the public conversation about this change has concentrated on one danger: that people will be fooled.

This conceptual, non-empirical paper argues that the more serious danger lies elsewhere. Drawing on peer-reviewed scholarship published between 2023 and 2026 in social epistemology, disinformation and media-effects research, deepfake detection and political deepfake studies, including work on African elections, and research on provenance and content authentication, the paper distinguishes two harms. The deception harm is the familiar one, in which people come to believe something false because a fabricated recording looks real.

The scepticism harm is the loss of belief in things that are true, because the mere possibility of fabrication gives every viewer a reason to doubt and every exposed actor an excuse. The synthesis suggests that the scepticism harm is larger and harder to remedy. It is produced by general awareness that fakes exist rather than by any particular fake, it survives correction, and it is actively exploited by those who dismiss authentic evidence against them, a pattern known as the liar's dividend.

From this distinction the paper develops an authenticity-infrastructure framework, which holds that a public's epistemic health depends less on the ability to detect fakes than on the availability of provenance, attestation by accountable parties, and trusted intermediaries that can establish what is real. Six propositions specify the conditions under which publics keep the capacity to agree on facts. The paper then examines conditions in Africa that heighten exposure to the scepticism harm, namely low institutional trust, the circulation of political content in closed messaging groups, and limited platform investment in content integrity for African languages and markets.

It concludes with recommendations for newsrooms, electoral bodies, regulators and platforms.