Evidence as content

Source: Rob Horning, Internal Exile · 2026-04-16

Not only that, but the tropes that make images seem “evidentiary” or “indexical” or “documentary” can also be optimized for and slopified, extracted algorithmically from billions of images to present the feeling “this is real” as a generic, malleable quality, as something simultaneously very clearly fake. This makes for a kind of meta-experience in which we can suspend disbelief in our ability to suspend disbelief. At the same time, the feed’s ceaseless demand for novel images requires that simulated “news” images be constantly produced and just as readily dismissed, scrolled through, evoking topics being discussed generally and shaping images that can be directed along different vectors toward different ad hoc demographics derived from troves of user data.

Like all newsworthy events now, the current Middle East war has brought about “a cascade of AI fakes” that have filled the feeds of social platforms. The nations at war may be conducting an all-out information blitz across any available media, with no referees or monitors to constrain it, but at the same time there is also a free-for-all race to use the same weapons to pander to audiences across the spectrum. Sometimes the clips serve explicitly propagandistic aims, hoping to melt enemy morale or forward a false sense of what happened, but often they are merely opportunistic, giving platforms material to stick in a user’s feed so that every time they refresh it they find something new and interesting, unhampered by any of the lags in traditional verified reporting or the constraints of needing to find genuine sources and illustrative details. Such clips cater to an audience that might vaguely think it wants to “monitor the situation” but is also possibly just bored and hoping their phone will keep them amused.

Regardless of their overt political slant, such clips are indifferent to whether they inform or misinform us, as long as they hold our measurable attention. They are simply propaganda for themselves — for the idea that all news can and should ultimately be experienced as entertainment. On-demand footage should exist of anything that might fit with any user’s expectations of what they should be seeing, and not necessarily what has actually occurred. Everyone should see content that rewards the tiniest amount of attention with a comforting sense of having already gotten the point.

Cumulatively, these clips don’t add up to an overarching narrative, even a false one. Instead, they encourage a post-narrative mode where enjoying memes doesn’t depend on understanding the complexities of the bigger picture or getting more context. Instead news footage, simulated or real, is treated as freely combinable with other material from the cultural archive, allowing us to enjoy spotting the formulas and references with no heed to propriety or decorum. When the White House mixes war footage with scenes from action movies, video-game screengrabs, or pro sports highlights, it is seizing on this post-narrative approach, which corresponds to Trump’s indifference to coherently explaining why he thinks the war needs to be fought. Like the memes, it is self-explanatory; you get it or you don’t.

But why would any actual person want to watch something fake? This might seem like a natural question if you think of social media feeds as news, but it’s silly to ask about entertainment. We engage with various forms of fiction all the time, in part because we like wish fulfillments, escapism, the feeling of suspending disbelief, the apparent confirmation of our beliefs and biases. It can be satisfying when something we maybe couldn’t imagine so vividly looks just like we would have expected, just like we’d want it to. These clips aren’t trying to justify themselves through being persuasive but by being diverting, by subordinating the complexity and overwhelmingness of events to our personal response. It’s more important for them to be emotionally legible than genuine — graspable in an instant.

AI slop clips often latch on to news stories and try to show viewers what the news would be like if it could be stripped of its context and presented as maximally engaging to an individual user. In this clip, of a dog pulling a stroller out of oncoming traffic, the caption asserts that what’s shown is “similar to real events that have taken place,” as if on the implicit understanding that an approximation is not only sufficient but perhaps preferable. Why settle for a real event when you can consume an enhanced and heightened reality presented in a similar idiom?

Both of these clips, purported to be filmed by mounted surveillance cameras, depend on an idea of what viewers expect visual evidence to look like, but convert that expectation into a purely formal component, something that is just another trope to play with. Like the war clips, they are not so much indifferent to facts as parasitical on the means of establishing them, illustrating how evidence can be produced as content — that is, media shaped by the incentives of social media feeds — and ultimately as slop: low-effort content whose production can be increasingly automated.

Evidence establishes the factuality of something regardless of whether people want to pay attention to it or believe it; content simply caters to and reinforces the desire for spectacle. As evidence and content have come to co-mingle in feeds, it has created a confusion between the two. Evidence-like material becomes an alibi for spectacle; it mimics verifiability but invites us to disavow any need for verification. We can experience what we see as true without having to fit it into a narrative or a confirming context. Like a magic trick, it temporarily short-circuits skepticism to allow for a kind of pleasure in simply accepting what we perceive. They also reinforce the feeling that it is more fun to be tricked than informed, and that indifference to truth yields instant emotional rewards.

Sometimes, as with magic, generated clips portray miraculous or catastrophic or impossible events, catering to escapism. But they also show what we expect to see. Slop caters to an idea that anything should be viewable, especially in a world so saturated with surveillance as ours: Hasn’t it already reached omnipresence, and shouldn’t omnipotence follow? This corresponds with the idea that generative models already contain everything that could possibly be perceived in its matrix of billions of data parameters. Slop clips deliver on both fantasies at once: AI can show you anything, and it is truer than ordinary truth.

These images proliferate not because they are mistaken as authentic documents but because, as Claire Wilmot argued in a piece for the London Review of Books, they “offer clear, illustrative diagnoses” of “alleged problems” — they are “showing everyone what’s really happening,” as someone knowingly sharing anti-immigrant deepfakes tells her. The clips don’t offer factual information — which would underscore our inability to intervene in the large-scale events immiserating us — but ideology, saving us the effort of having to think things through.

Generated video caters to an audience’s desire to see their beliefs in the form of evidence, and consume what they already believe as if it is being irrefutably established by real events. They visualize simple explanations for us so that we don’t have to put effort into them or take responsibility for them. These kinds of clips allow us to experience such beliefs as content without having to “really” believe in them. They just appear before our eyes to provide reassurance. They are self-verifying for their audience because they depict an idea that they already take for granted.

The steady flow of such material normalizes the idea that the news will immediately be recast into more satisfying shapes, and that we can depend on our feeds to allow us to modulate our emotions and be fed a kind of clarity, if not relief — a manageable rhythm of disbelief repeatedly resuspending itself. The flow of images, and the emotional regulation it affords, outweighs the impact of any particular image of video.

This is an underlying message of all content: that we should be able to see what we want to believe without having to invest effort into building that reality. Slop may have its social uses —animating unimaginative group chats, perhaps — but mainly it invites us to isolate ourselves. Rather than seeking out information from a recognition of our personal ignorance and out of a sense of civic responsibility — to help build and participate in a political world shared by others, grounded in ideas that have been collectively developed and articulated — we can ingest massive quantities of simulated information and clumsy propaganda that makes us feel full of ourselves. If civic duty as a purpose for becoming informed has atrophied; slop circulation aims to obliterate it entirely.