Kick off the season by testing The Grand Continent. The ideas and analyses you won’t find anywhere else. But you’ll see them everywhere soon. In your inbox from eight euros per month
During the summer of 2025, my social media feed was flooded with images of unruly crowds heading toward European ports and squares to loot and vandalize.
They are genuine barbarian waves battering the charming Navigli of Milan, the fashion capital, reducing them to heaps of trash. They storm the famous Ramblas of Barcelona and scatter waste across Europe, from Liverpool to Berlin via Copenhagen.
None of this is of course “real”: these are videos generated by artificial intelligence that claim to offer a glimpse of life in European cities in 2050. These clips, produced according to the rules of the short formats that are immensely popular on social networks, draw on the genre of time travel filmed in a POV (first-person point of view).
These contents reproduce a recognizable and easily reproducible visual grammar: waking up in ancient Rome as a gladiator, within the Vatican dressed in the robes of the pope, or in future Paris, these images invite you to “live” history in the first person, as a participant rather than a spectator.
It doesn’t matter that this narrative is entirely fictitious, whether set in the past or imagined in the future: it is a kind of fast history produced to grab attention and rack up likes rather than to document or inform.
Creators defend themselves by claiming that their aim is to spark public curiosity and push viewers to go further. Yet it is precisely in their openly speculative nature that the danger lies.
These videos are hardly mistaken for “real” imagery. Aesthetically and narratively, they are openly presented as fictions that reimagine the past or imagine alternate scenarios, present or future. They do not seek to deceive, unlike deep fakes. They instead give a tangible form to fantasies and possibilities, bringing them closer to our perceptual horizon, making them familiar and, gradually, conceivable.
The POV aesthetic further intensifies the feeling of “being there,” of experiencing a testimonial from within. Whether it is 1922 Italy, imagined with all the stereotypes of the dolce vita—the attentive wife serving breakfast in bed, the soccer match among friends on a grand monumental square where coffee and ice cream are served—or 2050, overrun by hordes of migrants who turn Italian beauty into a triumph of chaos, litter, veiled women to the eyes and bearded men in supplicatory prayer to Islam, the mechanism remains the same.
The complexity of a historical period is reduced to an emotional tone, “in the style of,” according to the now familiar formula used to prompt AI to generate images: nostalgia expressed as aesthetics, the visual idealization of a year, 1922, coinciding with the rise of fascism in Italy and the beginning of its disastrous consequences that we all know from factual history. But if the past is reconstructed as an aesthetic ambiance, the future is even more troubling. Conspiracy theories such as the Great Replacement or fantasies about Europe’s future submission to an ill-defined Islamic domination — the “dhimmitude” theorized by Bat Ye’or, pseudonym of Gisèle Littman, in works like Eurabia — acquire, with AI, an immediate and viral visual form, further reinforced by immersive formats.
How AI Makes the Memories of the Future
When I first saw the videos documenting the events that unfolded in Ceuta on 30 and 31 July, with the tragic arrival of more than 70,000 migrants on the shores of the Spanish enclave, these images felt strangely familiar. I experienced a sense of déjà vu, as if I had already witnessed these events, as if those scenes had already flashed before my eyes. Then I remembered the POV videos from 2025.
Hordes of people literally storming the coast, waste scattered on the beach and in the town’s squares: the visual construction of a true assault, a take, accompanied in many cases by soundtracks and a language evoking a siege. Watching and rewatching these factual videos from the summer of 2026, I understood what had happened to my gaze the summer before. The subjective videos had functioned as a kind of visual training. They did not convince me of their reality, but they silently reshaped my view, conditioning my understanding of events, making them familiar and thus recognizable when they appeared as factual images. Like a visual prophecy waiting to be validated by fragments of empirical reality. It is “normal,” it is “natural” that it ends this way: sooner or later, migrants will storm Europe.
This is what one could call speculative violence: a subtle form that nests in the folds of AI-generated images, precisely those that do not claim to be true and that, for this reason, can circulate unimpeded on the Internet and social networks. These are images that give visual form to problematic worlds and concepts, such as the Great Replacement, remigration, and other supremacist and xenophobic imaginaries. But once translated into the almost naive aesthetics of AI style, they spread as parodies, jokes, mockery, and seemingly harmless content. Until they become “real,” that is, until something they have been constantly prefiguring bursts into reality, validating the image and making it iconic and prophetic.
The new regime of looking, which far-right movements around the world seem to have learned to master with unusual efficiency, is not interested in representing reality but in manufacturing its strategic anticipation.
That is where the real game is played: it is not about showing the world as it is, but about making certain configurations visible by entrusting them to the most contemporary and ubiquitous aesthetic forms, from synthetic photorealism to AI’s slop, and by familiarizing the gaze to render their realization, in a sense, inevitable.
Speculative violence operates precisely in this interval: it may not necessarily distort the present, but it prepares the collective gaze to recognize and accept a possible future among others. It emerges where imagination is used to prepare for a violent project involving the expropriation, deportation, or erasure of populations or territories, and thus becomes a true instrument of governance.
The Prophetic Style
The AI-generated video shared by Donald Trump in February 2025, which showed the Gaza Strip devastated by war being transformed into a luxurious seaside resort, is a paradigmatic example. In these images, there is no blood, no rubble, no explicit destruction. On the contrary, the visual frame mobilizes the seductive language of prosperity and urban regeneration, presenting a fantasy of peace realized through real estate investments. The violence does not appear anywhere, yet it underpins the entire scenario. Beneath the smooth surface of these images lurks an implicit foundation of genocide, ethnic cleansing, and forced deportation: this bright future presupposes the physical or geographic elimination of a large portion of Gaza’s population, a notion Trump himself has repeated and is even more explicitly voiced by Benjamin Netanyahu.
When the “Gaza Riviera” project video began to circulate, sparking widespread indignation, it was described as grotesque satire, an orientalist fantasy, and yet another provocation in the now familiar Trumpian style. Yet, when the “20-Point Gaza Peace Plan” was officially presented and approved in October 2025 by an international coalition, few people noticed the similarities.
Restated in language that appeared more sober and bureaucratic, its content did not stray far from the AI-generated vision: peace was portrayed as an opportunity for real estate investment and profit, depriving Palestinians of any say in the future of their land and their lives.
What just a short time before had seemed scandalous had quietly crossed—the boundary separating speculative imagination from political reality—through repetition and normalization. This does not mean, however, that generative AI is the cause of the Gaza Peace Plan in 20 Points, nor that the images automatically translate into on-the-ground policies.
The same goes for Ceuta. POV videos do not fulfill the openly propagandistic function of deep fakes, which falsify reality or seek to replace it. They anticipate it: they quietly construct certain possible configurations and present them as speculative visions, and thus as apparently harmless fictions that are difficult to classify as propaganda. We still do not know which evidence Spanish Prime Minister Sánchez refers to when he accuses Israel or Russia of helping amplify the Ceuta crisis through disinformation. What we do know is that with these images of speculative violence, we are confronted with something subtler and more sophisticated than fake news, precisely because these images do not immediately register in the factual register. They do not inform that something happened when it did not; they show us what could happen, acclimating us to recognize this vision long before it arrives in reality.
If it is difficult to trace the real identity and political affiliations of their authors, their ideological orientation becomes clear.
From this universe of synthetic history emerges the construction of a composite enemy, capable of condensing into a single figure elements often incompatible with each other, such as the defense of LGBTQ+ rights and ultraconservative Islam, augmented by pro-Palestinian activism and the absolute dictatorship of veganism. In one of these Europe-of-tomorrow videos, for example, one can see a Putin-less Moscow that has tragically Islamized, while a Zelensky-less Ukraine is imagined as prosperous, peaceful, and democratic.
In the face of this threat, the parties supposedly tasked with saving “European civilization” appear to take on the traits of a nationalist bloc, pro-Putin and pro-Netanyahu, akin to MAGA. It is no accident that the protagonists in these videos often wear T-shirts bearing the slogan “Make America Great Again,” as if by analogy to evoke a supposed lost European grandeur: an imagined continent, prior to mass migrations and multiculturalism, whose integrity would now be under constant threat. It is that same mythified Europe that resurfaces in the Trumpian imagination, from Elon Musk to Marco Rubio: not a place that truly existed, but a past grandeur under siege by Islam and migration.
To stem the imminent barbarization of the continent, the proposal is to restore supposed European values, visually suggested through the AI’s synthetic imagination. One of these videos, titled You wake up in Belgrade in 2050, effectively encapsulates the recipe for surviving the dreaded Muslim conquest, offering a kind of visual manual for resisting alterity. It shows a migrant and asylum seeker detention center, a church, emblem of the persistence of Christian identity, a man eating a pork sandwich, and finally, a perfectly orderly and pristine public square. In this seemingly banal sequence of images, border control, Christianity, the consumption of pork, and urban cleanliness are elevated to the pillars of a besieged civilization whose survival would depend on defending its borders—territorial, religious, cultural, and even dietary—from an otherness perceived as a perpetual threat.
That “visual conservatism,” which manifests in Belgrade’s POV through explicit symbols, can take much subtler and apparently apolitical forms. Last year, the daily Haaretz noted significant aesthetic similarities—and thus value alignments—between some content used to influence Romania’s 2024 elections and items meant for Israel’s October elections. In both cases, AI photorealism is used to construct scenes of idealized daily life—birthdays, newborns, freshly harvested foods—that seem unrelated to politics.
And yet it is precisely this apparent innocence that makes it powerful. Acting as true visual lures, these contents would lure unsuspecting users into Facebook or WhatsApp groups, gradually steering them toward networks and narratives that are explicitly political, linked to the agendas of what the article calls “foreign actors,” among whom Russia figures.
If such operations are still, at least in part, demystifiable—as shown by the Romanian case, where the first round of the 2024 presidential elections was annulled, notably on the grounds of foreign interference via digital tools—what will become of speculative-violence imagery? In those cases, it is still possible to identify fake profiles, reconstruct influence networks, and perhaps assign responsibility.
Images Do Not Need to Be True
But what could we accuse POVs of regarding Europe’s barbarian invasions? These videos remain on YouTube channels, on Instagram Reels, scattered across the Web. They do not falsify an event that actually occurred, they do not forge someone’s identity, and they do not claim to document reality. How, then, should we formulate a precise accusation about the déjà-vu these images can instill in a viewer who, months or years later, faces factual images that suddenly seem to confirm them?
It is precisely in this ambiguity, in this liminal space where images prepare to anticipate the future—and, through contemporary, captivating, and instantly shareable visual forms, to suggest it as plausible, even inevitable—that speculative violence thrives. It is precisely because they are difficult to detect by traditional verification, censorship, or moderation that these images can gain visibility and permanence, embedding themselves in the digital landscape and contributing to remold how visual knowledge—and by extension memory—are produced and preserved.
The synthetic visual regime opened by generative AI is one in which plausibility, repetition, and virality can become criteria of legitimacy more powerful than truth. But the essential point is this: images of speculative violence do not need to pretend to be real, because they do not operate on a binary true/false logic. They do not ask to be believed. They ask only to be seen, repeated, and remembered.
How can we forget synthetic images like those of the Colosseum, where leaders of the Italian left are killed by European Parliament member Vannacci, who pretends to be a Roman emperor? We have all seen them and have all helped circulate them: some with disgust and indignation, others with approval and applause, and to illustrate the absurdity to which our civilization seems to have arrived—forgetting that the same images, which provoke laughter, also consume water, energy, and other resources necessary to produce them. “This is unreal AI-created satire,” retorted Vannacci with disdain when asked to distance himself from the video, contrasting it with the “real violence,” which, in his view, would be the act of the left, referring to clashes in Bologna following the demonstration after Abderrahim Fakir’s death.
By relegating the Colosseum images to the realm of unreality, Vannacci can dodge accusation: it is only fiction, satire, AI. But in the visual and synthetic ecosystem that is developing through the mass diffusion of generative AI, can we still rely on the factual/fictional, true/false categories? In a hyper-visual and accelerated contemporary culture, where virality and circulation create visibility without necessarily producing truth, these images function as performative projects, operational horizons, weapons of a war waged through visual speculation about the future, both aesthetically and semantically as well as politically and financially. They do not need to be true to yield effects in the real world.