Occupying the Imaginary

In February 2025, American President Donald Trump circulated an AI-generated video depicting Gaza as a grotesque mega-riviera.1 Monumental golden statues of himself littered a landscape of luxury hotels, palm-lined boulevards, and beachfront developments, while Benjamin Netanyahu appeared lounging in the sun, replacing a land and its people still undergoing systematic obliteration with a vision of colonial redevelopment.

From the discourse online, the video was widely dismissed as another instance of ragebait in its purest form. Yet the ease with which this AI slop invites ridicule risks overlooking the political worlds to which it unceremoniously gives visual form. A year later, elements of this speculative vision resurfaced in Trump's own hallucinatory reconstruction proposals for Gaza at the World Economic Forum.2 The slop-image not only imagined this version of a future for Gaza, it contributed to making one thinkable and executable.

Recent scholarship increasingly understands generative AI not only as a vehicle for mis- or disinformation but as a new digital infrastructure for propaganda. Generative models fundamentally reorganise how propaganda is produced by dramatically reducing the cost of creating ideological content while enabling its rapid dissemination, endless variation, and mass personalisation across audiences, contexts, and reactive moments. Propaganda mutates from a fixed message model to a dynamic system capable of continuously generating virtual political worlds at unprecedented speed and scale.3

The term 'AI slop' refers not just to the degraded presentation of these images, but also to the impoverished ideological propositions they make. Low resolution, repetitive compositions, anatomical inconsistencies, and garish colours are the constitutive aesthetic features of how AI slop communicates. Slop does not fail to achieve plausible documentary realism; it largely abandons it altogether, and thrives through this aesthetic of convoluted excess.

Crucially, the political significance of these images rarely depends on whether they are mistaken for something ‘real’, as in documentary evidence.4 Often, they do not even bother pretending. Rather than persuading through isolated acts of deception, synthetic media manipulates well-established cognitive mechanisms like negativity and confirmation bias to promote its own circulation.5 Generative AI amplifies these tendencies by making it possible to continuously microtarget audiences with emotionally resonant content tailored to their existing dispositions, effectively flooding the zone with endless variations of the same ideological narrative.6

While the most visible examples of slopaganda have emerged from the American right, largely because of Trump’s own embracing of AI-generated political spectacle, one of its clearest regional offshoots was the now-defunct Austria-based YouTube channel MLGA NOW (‘Make Lebanon Great Again’).7 Launched in March 2025, the channel spent over a year producing and aggressively promoting AI-generated videos imagining a post–Hezbollah Lebanon amid the country’s war with Israel, forcing them onto Lebanese social-media feeds through paid advertising. Borrowing both the slogan and political ethos of Trump’s movement, its videos featured President Joseph Aoun and Prime Minister Nawaf Salam addressing the Lebanese public directly through clunky AI-generated counterparts. Together, they parroted a familiar set of talking points: Hezbollah must be disarmed, the Lebanese Armed Forces are the country’s only legitimate military institution, and Lebanon must pursue normalisation with Israel as a segue into permanent peace between both countries.

Towards the end of the channel’s life, and during the US-Israel war on Iran, this gave way to a more ostensibly centrist position in which both Israel and Iran should burn in favour of Lebanon’s alliance with the United States. The ideological throughline was often difficult to locate beyond pro-US and anti-Hezbollah sentiment, but that was largely irrelevant. Despite their frequently absurd, even parodic quality, the strangely affectless tone was fitting of the politics on display. The channel had garnered over 75 million views by March 2026; a flooding of the zone par excellence, from overseas nonetheless.

Much of the existing literature on slopaganda focuses on elections and other moments of electoral contestation. The above examples present a sharper and more exceptional case. MLGA NOW emerged in the midst of Lebanon’s ongoing war with Israel, as the newly elected government negotiated with the United States over an Israeli withdrawal from southern Lebanon in exchange for the disarmament of Hezbollah. The content was a response to this political moment, seeking ‘participation’ in it by manufacturing a dubiously utopian vision of Lebanon. 

At the same time, this barrage of mythical discourse and accompanying aesthetics escaped the confines of the Lebano-Trump political panopticon and took on a broader cultural life. TikTok accounts such as @sigma.leb (also known as Christian Maxxing LB) were preoccupied with reviving older far-right Christian imaginaries, circulating AI-generated images of Bachir Gemayel strolling through present-day Beirut with Raouché Rock in the background, before reimagining him as the smiling patron of a Formula One circuit dressed head-to-toe in the Lebanese flag.8 Simultaneously, accounts self-branded as history pages frankensteined endless Phoenician fantasies depicting Lebanon as a timeless Mediterranean civilisation detached from its contemporary political entanglements, complete with an imagined Phoenician passport.9 More recently, similar accounts have begun imagining a Beirut of 2040, complete with futuristic skylines, World Cup stadiums, and revitalised infrastructure, all presented as the natural outcome of a Lebanon that made the ‘right’ political decisions.10 Even MLGA merchandise is now pushed through AI-generated Instagram advertisements, announcing that ‘the war is over’ and that ‘now it’s time to build’, mirroring the rhetoric surrounding the US-brokered negotiations between Lebanon and Israel.11 Although presented as a pathway to peace, or at least a permanent ceasefire, the agreement was criticised within Lebanon for conditioning Israel's withdrawal on Hezbollah’s disarmament rather than setting a clear timetable for ending the occupation. Critics argued that the arrangement risked normalising a devastating open-ended settlement that would facilitate and normalise Israeli presence in southern Lebanon.12

This is not to suggest that there is any political coordination between these accounts or that they even share the same political agenda. What they do share, however, is their participation in building the same semiotic universe, one whose formula is remarkably simple and where affect increasingly precedes argument or sustained political thought. By endlessly recombining a handful of immediately legible signifiers – Bachir Gemayel, Phoenician iconography, the Raouché Rock, FIFA stadiums (vis-à-vis the country’s famously neglected infrastructure), and futuristic seaports (vis-à-vis the barely functioning Port of Beirut after the 2020 explosion) – these images distil decades of political turmoil into an immediately recognisable fantasy of national redemption, one increasingly detached from the material realities on the ground.

Viewing scalability and personalisation as the defining novelty of generative AI in propaganda holds its own merit, but it reveals comparatively little about the images themselves. If generative AI is in the business of assembling its worlds by recombining statistical patterns from existing image archives, then the archive itself becomes just as politically consequential as the propaganda it generates.

Like Trump’s Gaza riviera, these images inundate the collective imagination with generated futures (and pasts and presents) assembled from unresolved historical desires. They ask very little of the viewer beyond immediate recognition, provoking feelings of nostalgia, grievance, rage, and other dopamine-inducing affective states. In doing so, they saturate the political imaginary with endlessly reproducible myths of the nation.

These generated worlds appear speculative on the surface, but they remain fundamentally retrospective. Generative AI models assemble imaginable visual worlds from patterns extracted from billions of pre-existing images. Every output is built from accumulated visual conventions carrying with them the cultural assumptions and ideological embeddings and biases of the archive itself. What appears as an image of the future is, in fact, a statistical reconstruction of an already mediated past. The machine’s dreamed-up futures are therefore acts of historical selection by design.13

The implications of this operation become apparent once these image-making systems are mobilised in the service of propaganda via AI slop. Trump’s vision of Gaza marries the language of luxury developments, emptied landscapes, and spectacular reconstruction with a much-longer visual history of colonial modernisation. In the same vein, the AI-generated Bachir Gemayels, futuristic Beiruts, and self-affirming Phoenician myths reactivate unresolved imaginaries and dream visions already lodged in the Lebanese collective unconscious. 

The result is a gradual corrosion of the political imaginary, where visions of prosperity displace the difficult work of imagining practical solutions to Lebanon’s problems that are rooted in the country’s geopolitical reality. The repeated return to Phoenicianism, for example, is a recurring fantasy of civilisational exceptionalism. It offers a symbolic distancing from Arabness by imagining Lebanon as exempt from a regional political condition dictated by the presence of a genocidal state to its south, reactivating one of the constitutive fantasies of the Lebanese state: that the country is historically and civilisationally distinct from its Arab surroundings.14

It is unsurprising that AI imagery has increasingly been dubbed as the aesthetic of fascism.15 The phrase is an observation about the kinds of political narratives it appears particularly adept at bringing to life. Across Lebanese social media, AI-generated fantasies of normalisation with Israel and subsequent prosperity are ample.16 The complexities of history and political conflict dissolve into the slop, giving way to spectacle.

The machine therefore performs a curious inversion. It presents itself as a technology of unprecedented innovation while repeatedly returning to long-standing political fantasies. This occurs even before accounting for the opaque systems of training, alignment, and moderation that further inscribe particular political assumptions into different models.17 These images do not illustrate an ideology after the fact; they mobilise slop as a visual form to stage a virtual political horizon in which selected political fantasies are rendered present, familiar, and cognitively available.

The political orientation may change, but the slop-image remains remarkably consistent. Accounts such as @thewartoon, where Iranian drones, missile systems, and other military hardware explain their own function through cutesified, AI-generated animations, demonstrate just how readily malleable it is.18 Only a year ago, most people could not have told you what a Shahed-136 drone was, let alone have it explain its own function to them in the form of an AI cartoon. The US-Israeli war on Iran soon became another proving ground for this visual language. Pro-Iran media pumped out AI-generated, Lego-style videos depicting successful Iranian military retaliation on the US, which circulated widely online, adopting the same slop aesthetics to narrate the conflict from an opposing geopolitical position.19

Whether deployed to imagine a post-Hezbollah Lebanon, celebrate Iranian retaliation on the United States and their allies in the region, or explain contemporary weapons systems, the same aesthetics repeatedly transform war into an easily consumed visual meme. The same image is informing, entertaining, and propagandising all at once. These virtual images of warfare thus accompany, anticipate, and participate in the war itself.

As ideologically charged images become cheaper to generate and easier to circulate, the concern over the conditions under which political reality is perceived is well-placed. The slopaganda at hand does not always necessarily argue for particular outcomes – it reorganises perception around selected and hallucinated worlds, making them visually familiar before their physical manifestation becomes possible. The battlefield is not confined to territory or information alone: it is mediated through the management of what can be seen, heard, thought, and felt, where synthetic images inculcate political ideologies and subsequent settlements before they can be contested.20

The distinction between imaginary visions of the nation, where the figureheads of the Lebanese government happily pose beside Donald Trump, all wearing MLGA caps, and material warfare is, however, not that stark. They both share the same infrastructures that govern military action on the ground. This overlap manifests in Israel’s well-known expanding use of AI-assisted military systems throughout its wars on Gaza and Lebanon. Platforms such as Lavender, The Gospel, and Where's Daddy? have been widely reported as assisting in identifying, prioritising, and targeting thousands of individuals and civilian infrastructure.21 The same systems that automate the ideological flooding of social-media platforms also automate the violence through which those imaginaries are pursued.

It is easy to dismiss AI slop as beneath serious political analysis. After all, who is actually fooled by these images? Surely the natural instinct of most users is to ignore or scroll past this assault on the senses? But this posture mistakes the conditions under which propaganda has always operated. The image has never needed to be convincing or pleasurable to be politically useful. Propaganda has long depended on priming individuals through the gradual and repeated exposure to particular ways of seeing the world. AI slop lowers the threshold for participating in the symbolic battlefield even as the means that make such participation possible become concentrated in the hands of a small number of Big Tech companies and their oligarch CEOs.

None of this should read as technological determinism or an invitation for more moral panic around generative AI. But what distinguishes this moment is the remarkable ease with which political worlds can now be generated and amplified at scale. What is new is not the fantasy itself, but how little it now takes to wage and participate in a semiotic battle. When battles are fought as much in the imagination as on the ground, occupying territory requires occupying the imagination first.