After Evidence

On January 1, 2026, Adam Mosseri, head of Instagram, published a year-end reflection on the platform: a visual and textual essay examining how artificial intelligence is radically reshaping the way we interpret images and videos online. In his statement, he makes a clear assertion: the era in which we could assume what we see is real has come to an end. We now operate within an ecosystem where AI-generated content has reached a level of sophistication that renders it indistinguishable from authentic media.

The proposal that emerges is sharp and, in many ways, unsettling. Rather than attempting to chase down and expose every individual fake, it becomes more rational to shift the problem upstream, toward the certification of reality itself. No longer an endless hunt for falsification, but the construction of credibility systems, reliable author-based signals, and mechanisms for authentication at the point of media origin. In other words, the issue is not about saving the image, but about redefining the conditions that allow it to carry weight.

We have not entered an era in which images lie. We have entered an era in which images are no longer able to declare their own nature. The distinction between the real and the artificial has not merely thinned; it has dissolved as an operational criterion. Asking whether an image was captured or generated no longer produces knowledge, but rather a permanent suspension of interpretation.

Indistinguishability is not a temporary accident caused by still-imperfect models. It is the stable condition of a visual ecosystem in which the capacity to simulate reality has reached a systemic threshold. When simulation no longer betrays itself, perception loses its role as arbiter. Not because it fails, but because it no longer has parameters on which to operate.

This shift produces a profound effect: vision ceases to function as an act of verification. Seeing no longer equates to knowing. The eye, which for centuries served as an interface between the world and evidence, becomes a simple access point to equivalent surfaces. The real image and the synthetic one occupy the same cognitive space, with the same persuasive force and the same formal density.

In this context, the question “is it real or is it AI?” quickly becomes sterile. Not because it is wrong, but because it arrives too late. Indistinguishability operates prior to judgment, upstream of analysis. It is a pre-critical condition that neutralizes distinction before it can even be articulated. Visual culture does not fail; it changes regime.

What emerges is not a crisis of the image, but a crisis of the automatic trust we once attributed to it. The image does not lose value because it is artificial; it loses value because it can no longer guarantee its own origin. In the absence of this guarantee, every visual content becomes a claim that demands something else: context, source, history, responsibility.

We might argue that indistinguishability is the new ground on which contemporary visual communication is built. And like any new ground, it requires different tools: no longer based on evidence, but on provenance. No longer on seeing, but on the ability to reconstruct. In this article, we will attempt to raise a series of questions around this condition.

For a long time, the image functioned as a cognitive shortcut. To photograph, film, or record meant producing a trace of reality considered reliable enough to require no further mediation. The image was both representation and document. It did not ask to be interpreted; it presented itself as proof, almost by default.

The arrival of artificial intelligence does not destroy this function; it simply renders it unworkable. When the same visual quality can be achieved without any contact with the world, the image ceases to attest to an event and begins to behave like a statement. It says something, but it proves nothing. This is where the rupture occurs: the image loses its evidentiary status.

This does not mean that real images disappear, or that everything becomes false. It means that their value is no longer intrinsic. A photograph no longer carries weight simply because it is a photograph, just as a video no longer does because it was filmed. The historical link between image and truth, already fragile but operational, is definitively severed. From that moment on, every image is guilty until proven otherwise.

The paradox is that this loss of evidential force makes the world appear more uniform. Everything looks plausible. Everything seems possible. And it is precisely this hyper-plausibility that empties the image of its power. When every surface can be produced, the surface no longer distinguishes. It does not orient, at least not yet.

In this new regime, the image does not fail because it lies, but because it can no longer guarantee its own origin. It is no longer able to say, “I come from there.” Without this indication, vision remains suspended: what we see may have happened, may never have existed, may be a synthesis of events, data, styles, archives.

The end of the image as evidence is not a collapse, but a change of function. The image shifts. From proof it becomes a surface of access. From document it becomes interface. It asks to be accompanied by something else, a source, a history, a system of attestation, in order to regain weight.

In an attempt to compensate for the loss of the image as evidence, digital culture gradually shifted its attention toward metadata. Technical information, invisible traces, timestamps, signatures, coordinates: elements designed to anchor content to an origin, a moment, a device. For years, they functioned as prosthetics of trust, as if truth could be reconstructed through technical means.

Today, this balance has shifted as well. Metadata is no longer a foundation, but a secondary surface, fully integrated into the content itself. It can be produced, removed, or rewritten with the same ease as an image. Its presence no longer guarantees a link to reality; it merely signals the existence of a technical passage. Metadata becomes information, not attestation.

This shift has a precise consequence: trust can no longer be delegated to an invisible layer of the file. Not because technology has failed—quite the opposite—but because technology no longer coincides with responsibility. When the data meant to certify origin is as manipulable as the image it accompanies, the entire system loses its orienting capacity.

In this new scenario, metadata does not disappear. It changes function. From guarantee, it becomes clue. From proof, it becomes a weak signal, useful only when embedded within a broader structure. In isolation, it carries no weight. Contextualized, it contributes. Its value is no longer absolute, but relational.

The cultural consequence is significant: truth is no longer located within the file itself, but in the trajectory that generated it. Attention shifts from static content to the dynamics that sustain it. Who produced it, in what context, with what continuity, within which system of relations. Technical data alone is insufficient; it acquires meaning only when placed within a legible chain of responsibility.

The collapse of metadata as a foundation of trust does not create a void, but a realignment. Credibility is no longer automatically extracted from the structure of content; it is constructed through more complex and more costly processes. In an environment where everything can be generated, what matters is not what accompanies the image, but what endures over time around it.

At the moment when the image loses its ability to ground its own credibility, the visual system reorganizes its weights. Attention gradually shifts from the surface to the point of emission. It becomes almost inevitable to recognize that the source emerges as the new center of gravity of meaning—not as a simple marker of origin, but as a complex structure capable of sustaining content over time.

The source no longer coincides with a name or an account. It is a layered assemblage of elements: continuity, coherence, history, relationships, and the assumption of responsibility. These are factors that cannot be produced instantaneously and require time to accumulate density. In an ecosystem dominated by synthetic abundance, what cannot be immediately replicated gains value.

This shift produces a radical change in how meaning is assigned. The image is no longer read as an autonomous unit, but as a node within a network. Its force does not lie in what it shows, but in the source’s ability to render legible the context from which it emerges. The source does not guarantee the truth of the image; it guarantees the traceability of the process.

In this sense, the source becomes an active device. It does not certify reality in absolute terms, but establishes a threshold of reliability. It allows content to be evaluated according to the solidity of the system that produces it. Attention moves from “what do I see?” to “who stands behind what I see?”

This new configuration does not eliminate ambiguity. It makes it operable. Ambiguity is not erased, but framed. It is the source that provides the frame, not the image. And it is precisely within this frame that content regains weight—not because it is true in an ontological sense, but because it is embedded within a recognizable chain of responsibility.

The centrality of the source marks a return to a pre-digital logic within a radically post-digital context: when evidence is no longer available, trust is built over time, through the stability of a presence.

In the new visual ecosystem, credibility is grounded in the ability to assign weight to content. Weight does not coincide with visual impact, formal quality, or degrees of realism. It emerges when a piece of content is embedded within a structure that makes its origin, trajectory, and responsibility legible.

Certification assumes a precise role here: to make the genealogy of content explicit. To certify means to situate an image, a video, a file within a recognizable chain of production, publication, and relation. This chain does not guarantee ontological truth, but it constructs operational reliability. It allows orientation, evaluation, and the attribution of relevance.

The notion of weight is generated through resistance over time. Content acquires weight when it implies continuity, exposure, and the assumption of responsibility. Weight accumulates through coherent repetition, stable presence, and the possibility of being traced back to a source that endures. In an environment of instantaneous generation, time becomes the primary factor of differentiation.

Certification operates as a cultural infrastructure before it becomes a technical one. It organizes the relationships between content, sources, and contexts. It introduces a functional hierarchy based on traceability and relational density. Within this system, content does not matter for what it shows, but for the network of references that sustains it.

The central distinction thus shifts onto a new axis: traceable content versus opaque content. The former participates in a shared structure of meaning; the latter remains an isolated surface, easily replaceable. This distinction does not produce exclusion, but orientation. It allows visual culture to function even under conditions of extreme abundance.

The new infrastructure of credibility must operate, in our view, as a system for distributing weight. Not by assigning definitive certainties, but by establishing conditions of legibility. In a world where everything can be produced, what matters is what can be reconstructed. Value no longer resides in the image, but in the structure that makes it interpretable.