What if Donald Trump were the Pope? Jesus, a shrimp? Instead of mere noise, what happens when memes and AI slopme become the new frame of reference, guided by platform incentives and affective uptake? In times of global meme wars and generic AI, culture is once again presumed dead—and so is context. The endless recycling of online data takes on a world-making function. Dancing cats, copulating fruit, collapsing buildings, and longboarding soldiers all blur into a relentless stream of recommendations, detached from any personal preference.
Why Fakes Still Matter
AI slop feels personal. In a synthetic avalanche of moving images, where everything happens at once, and nothing really sticks, fake faces feel real because they remind us of other faces, and fake bodies move in familiar rhythms, like bodies in trending videos we have long since forgotten we watched. We may not remember, but platforms do – a brief replay, a glance at the comments, if you liked this, your “For You” page will display more of the same. What changes with AI slop is that “the same” is extracted from online culture at scale and pushed to absurd extremes in the pursuit of algorithmic gratification. The world to which AI images refer is grounded in a machinic reconfiguration of the information environment, where the collapse of the (im)probable and (in)authentic has become routine.
Why do fakes matter? We instantly know which images are artificially generated. Sometimes we’re uncertain. Other times, we simply don’t care. With generative AI outputs now native to social media feeds, fakes no longer require facts as a contrapoint. When something is prompted into being, shared, reacted to, remixed, or acted upon, it moves from potential to actual. In a recent AI slop session with the students attending one of my digital methods research seminars, I caught myself staring at a YouTube Short: a toddler running into her father's arms during a military parade, captioned “Daddy Soldier Comes Home”. American flags flutter in the background as the soldier struggles to hold back tears before finally kneeling in an embrace, “I gotcha, baby”. Right, I thought. Trump has just announced another war. AI-generated memes urging him to send Barron “to fight and die for Israel” are spreading anxiety and outrage. Here was yet another remix of the same script, swelling with synthetic affect. And still, for a brief moment, the video had caught me off guard.
Slightly annoyed, I watched it again, wondering why it worked. Was it because the clip’s choreography of patriotic ritual felt oddly familiar? Or because, seconds earlier, I had liked a video of a cat farting in a tub of soapy water, and my affective bearings were already compromised? I swiped the toddler away, and an Iranian soldier longboarded down a desert highway, drinking pomegranate juice as missiles streaked overhead. I swiped again, and three female soldiers posed before fighter jets, smiling: “Habibi, come to Iran”. The next swipe revealed a vending machine full of “adorable puppies for sale”. Certainly, AI slop is, as Jason Koebler suggests, a brute-force attack on the algorithms. Then again, like memes, clickbait, and porn, it also succeeds for a different reason: It vibes.
Vibes + Memes + Algorithms = AI Slop
AI slop vibes, so much so that a concept long at home in sound and affect studies now sits at the center of debates on platform culture. For Robin James, “vibes are vernacular versions of the methods algorithms use to perceive the world”. Vibes capture “how we perceive ourselves the way algorithms perceive us.” Steve Goodman in Sonic Warfare directly links vibes to the cultural workings of memes and brands, repeatedly drawing attention to their “vibrational effects.” Mitch Therieu describes how algorithmically curated platform environments help produce an ambiance “suffused with an overall vibe, a vaguely felt sense of aesthetic unity among diffuse and low-intensity sensations.”
In the introduction to the edited collection on memecry across platforms, produced for Platforms & Society, Marloes Geboers and I suggest that vibes are more than an aesthetic supplement to content – viral or memetic. As a means of ambient amplification, vibes shape how we navigate the diffuse elements – interactions, recommendations, interfaces, profiles, feeds – that platforms fold into a single, all-encompassing, engaging atmosphere. Vibes are affective states of effortless attention rendered consumable as a platform product. AI-driven meme wars, brainrot, and slopaganda – all recent symptoms of attention-driven platform ecologies – point to forms of appropriation grounded in atmospheric familiarity. Generative AI affordances, alongside social media affordances for recommending, remixing, and reposting, give rise to affinity-based algorithmic approximation. Artificiality is not all there is. But let’s take a step back.
Earlier algorithmic systems relied on explicit verification: if a condition was met, a specific action followed. Generative adversarial networks (GANs) transformed this conditional logic into neural network parameters, in which procedural distinctions between true and false emerged through the recursive interaction between the generator and the discriminator. The generator’s fake outputs succeeded insofar as they could not be distinguished from the authentic training data according to the discriminator’s learned criteria. Crucially, this authenticity test did not remain confined to the computational operations. Popular deepfake imaginaries similarly center on the capacity of synthetic media to deceive viewers or evade detection systems, reproducing the logic of verification at the level of cultural engagement. However, while GAN-era deepfakes presupposed the imperative of passing as real, today’s algorithmic environments shift the focus from verification to speculation.
With the rise of multimodal large language models (MLLMs), speculation extends beyond wagers on uncertain futures into the realm of generative co-creation. The explicit conditional “if…then” of machine learning – which, for Taina Bucher, operates as a strategy for reducing uncertainty through "predictions about the likelihoods of outcomes" – finds a new expression in stochastic interpolation. Unlike programming, natural language prompting does not follow predefined computational rules. Given an input prompt, genAI models estimate unknown values to create transitions between known data points. When we prompt for a "shrimp Jesus," AI finds coordinates such as "marine biology" and "Christian iconography" and interpolates between them. The system projects variations from learned distributions, transforming computation from the rule-bound execution of fixed conditions into the exploration of hypothetical orientations. Speculation, as a logic of trial and error, collapses into an endless series of “what ifs,” emerging from the recombination of heterogeneous and incomplete data assemblies.
This logic extends into the algorithmic infrastructures of social (and other) media platforms, from mood-driven recommender systems designed to anticipate what we are likely to engage with next to generative AI models embedded directly into the production and circulation of attention-grabbing content. Platforms organize visibility by foregrounding some forms of engagement and backgrounding others, exploiting this attentional architecture through algorithmic targeting of ambiguous affect. Whether the next recommended video is fake or real, the vibe good or bad, does not matter. The joys of hate-watching AI slop are by now well documented. What matters is the feed's capacity to generate enough intensity to keep us engaged. Your emoji defeated my argument, writes Jodi Dean in “Faces as Commons.” Slop is like that emoji. Whether optimized for outrage or fun, it fuels social media with competing attentional signals that outpace meaning.
How I Stopped Hating the Algorithm
While the proposition may appear overstated, it emerges in a context where “the algorithm” has become a familiar target of cultural critique and a convenient shorthand for the many systems that sort, classify, rank, recommend, generate, and predict. Consider dead internet theory, whose proponents interrogate AI – analytical and generative alike – for turning the web into a wasteland of commodified, non-human digital traffic. AI slop is a byproduct of systems trained to recognize and reproduce patterns already established within online attention economies. Some would argue it marks the exhaustion of online culture itself. Memes are dead. For Gregory Chatonsky, the prompt evacuates precisely the processual, exploratory, and dialogical dimension that characterized authentic memetic creativity. The context is dead. What remains is regurgitation of relatable synthetic material optimized for visibility through vectors or “directional intensities that constitute the algorithmized social body.”
Others would critically engage with pattern discrimination that masquerades as a seemingly neutral ‘common sense’. For Eryk Salvaggio and Roland Meyer, mainstream generative AI is predisposed to amplify conservative and reactionary visions. Their critique resonates with broader concerns about chatbots as bullshit machines, not because generative technologies deceive outright, but because they render historically contingent norms, stereotypes, and power relations natural and self-evident. GenAI models operate as bias engines, reproducing the ideological structures sedimented within their training data. They are structurally nostalgic in that they generate idealized imaginaries from cultural archives of the past. From this perspective, the flood of AI slop reflects a peculiar form of abundance: an ever-expanding supply of images drawn from a relatively stable repertoire of representational habits calibrated for short attention spans. The capacity of these systems to recycle racist and sexist stereotypes alongside attention-driven logics of advertising, pornography, and platform capitalism is virtually endless. Virtually, there are no limits, but as the web becomes progressively populated by AI slop, genAI models are expected to collapse, feeding on their own synthetic outputs.
For yet others, it is perhaps no surprise, then, that AI chatbots have come to be described as post-social and post-participatory media forms. Remarkably, such diagnoses even tend to echo the very qualities they attribute to AI itself: novelty, transience, and an alleged break with networked sociality. Petter Törnberg and Richard Rogers, in their recent preprint, argue that the generative AI revolution dissolves the social media paradigm by decoupling platforms from any dependence on human participation. At the same time, whether AI constitutes such a break depends largely on where one is looking from. From the standpoint of slop creators, the social dimensions of AI are difficult to ignore. The shared object of attention has become synthetic. It generates value.
Take Fruit Love Island. Every prompt is the outcome of collective experimentation refined through loops of successive feedback: creators reverse-engineer viral templates and combine multiple models via emerging studio platforms designed for cross-output optimization. Take X. Evil by design, a chatbot like xAI’s @Grok is baked directly into the platform’s connective infrastructure, extending prompt-sharing to wider networks of interaction through tagging in threads. Unlike individualized chatbot exchanges, Grok exchanges are public and traceable, as users continuously adapt prior human-machine co-created outputs. Another example is OpenAI’s failed attempt to become the new TikTok with Sora 2. It failed not because the app lacked networked publics, but because it was prohibitively expensive. Once your likeness can be re-prompted, modified, and memeified by others, the network needs to be maintained. And infrastructure costs. It costs massive amounts of human labor, data, and natural resources. The synthetic social is extractive. Some memes may indeed be dead, especially if we reduce them to mere pattern variations in user-generated content. Other memes, though, are better understood as a recursive interplay between user practices and algorithmic systems.
Analytically, this proposition comes with three observations. First, the synthetic is not reducible to the artificial, just as the social has never been merely human. Second, AI slop may be better understood as a speculative riff on emergent trends than a straightforward reaction to algorithmically inferred preferences. After all, no one doomscrolls through fruit porn wedged between war propaganda and dancing cats on their For You page, thinking: yes, this perfectly captures my taste. Finally, taste, whether personal or shared, does not correspond to how things make us feel. Social media algorithms, while purposefully engineered as engines of order, may increasingly amplify more ambiguous states of distracted attention, from which platforms extract value in the first place. Susanna Paasonen made this argument a decade ago, in her article about memes. Since then, not much has changed, except, as usual, the speed and the scale. For platforms, each recommendation is less a targeted prediction than a wager on affective intensity. For slop creators, each prompt variation is a speculation about the latent engaging potential in previous generative outputs. To study both, I realized I would have to stop hating the algorithm. To understand why slop vibes, I would have to embrace a sloppy research persona.
The Sloppy Research Persona
Let me continue with the premise that AI slop matters in its own strange, synthetic way, and that there is value in looking more closely at the dynamics by which cultural imaginaries become data, and data become affect-saturated media replicas that, in turn, generate further cycles of data-intensive engagement. Slopification captures this process, though not necessarily in the conventional sense of a decline in content quality. In terms of infrastructure, it refers to the way generative models learn from the sedimented traces of online attention, absorbing the dominant patterns of online culture. All media is training data. At the same time, it involves creators who rapidly rework these patterns in response to breaking events through outputs most compatible with the algorithmic tendency to reward immediacy and familiarity.
A research scenario that brings this premise into sharp focus is meme wars. Understood as online battles for user attention that cater to algorithmic visibility regimes, meme wars are a product of human-machine co-creation, in which cultural references, platform incentives, generative outputs, and user practices continually inform one another. Studying such processes requires a methodological sensibility responsive to their peculiar temporality, ambiance, and affective force. This, I suggest, entails three moves: first, inhabiting a research persona capable of engaging synthetic culture on its own terms; second, spending time with metadata surrounding recommended content; and third, accounting for the vibes.
For Liliana Bounegru, Melody Devries, and Esther Weltevrede, the research persona method is a speculative sense-making device. It involves producing an artifact to prompt an algorithmic personalization scenario that can be inhabited over time. The persona offers a way to examine personalized recommendation spaces, such as the For You feed, as a site of user figuration. My sloppy research persona is a slightly more promiscuous variation of this script. Meet elp-2026.
She is active on TikTok, YouTube, Instagram, and X. At her best, she is a lurker of a kind that Olga Goriunova describes as “not involved but performative.” She cares little for content specificity: it can be politics, current affairs, or entertainment. Variations mildly excite her. Her watch time and exposure patterns generate recommendation loops she may or may not endorse. She never follows. She never comments. Sometimes she likes slop. Sometimes she lets slop play on repeat. She pays attention only to what looks synthetic enough. In deliberately slopifying her feeds, she calibrates an affinity-based algorithmic space for future data extraction.
Each session with elp-2026 is a walk through a platform’s latent directional space. Each update is a temporary exposure of the vectors along which recommendation systems organize attention. In the context of the For You feed, a vector is a direction or computational orientation derived from accumulated signals such as watch time, pauses, replays, likes, and skips. As engagement becomes vectorized, content and users alike are transformed into multidirectional coordinates. The relations that emerge between them are operational rather than semantic. Entities that may appear extraneous can nevertheless be drawn into proximity because they occupy neighboring positions within a platform’s predictive architecture. In Chatonsky’s account of vectorization, this is how platforms optimize for content that triggers intense affective responses. Rather than unpacking this approach further in the abstract, we can explore it at work. What follows is a snapshot of elp-2026 on Instagram. For maximum effect, watch the video until you run out of patience. Notice what disappears, what repeats, and what feels unexpectedly connected. Content warning: things get sloppy.
For elp-2026, this approach raises methodological questions about how a sloppy research persona might account for her algorithmic encounters. What affinities surface with each new update? To what extent are content flows configured to resonate across the shifting registers of boredom, fascination, outrage, amusement, disgust, curiosity, and their many permutations? Each pause, replay, or like feeds forward into further recommendations designed to capture adjacent vibes. Her “For You” feed unfolds as a polyrhythmic constellation in which, as Dawn Lyon and Rebecca Coleman suggest, rhythm is not reducible to speed or sequence. Rather, it orchestrates the “relations, connections, and/or ‘interferences’ between the linear and the cyclical”, actively shaping user experience. As elp-2026 navigates from one recommendation to the next, interactions accumulate, and repetition becomes a mechanism of orientation. Each response to a familiar pattern reinforces the feed’s vectorial pull, bringing the next encounter closer to the previously expressed attentional tendencies.
In the Instagram session shown here, one such tendency is evident in the rhythm of Nigerian TikTok influencer Emeka Leonald’s iconic strut, which persists across synthetic adaptations featuring Trump and other political figures. Disrupting this rhythm are numerous competing directional vectors: military-themed synthetic thirst traps, uncanny celebrity-centered templates, and anthropomorphic animal memes staged in improbable scenarios. The sexualised accounts, in particular, appear as a predictable continuation of Instagram’s established attention economy, where bot farms and influencers alike have long exploited dominant visibility standards, now reproduced and amplified at even greater scale through generative media. Other platforms and use scenarios may amplify other synthetic vernaculars. Repetition is less a property of content than an invitation to think rhythmically through the feed, giving my research persona the scope to explore and express ‘the feel of algorithms’ that Minna Ruckenstein describes as constitutive of everyday online exchanges.
From here, the persona experiment opens different trajectories for interpretation. Conceptually, it offers a glimpse into the assembly of embodied performances and gestures from which algorithmic trends are continuously extracted, replicated, and recomposed. Methodologically, it foregrounds the need to reflect on the experiential messiness of such encounters as well as the need for elp-2026 to endure the multiple affective fluctuations that recommendation systems elicit through trial and error. Practically, it creates space for platform-native algorithmic environments to remain in a state of friction, where AI slop need not be reduced to manipulation or deception. Effectively, AI slop becomes approachable as more than one thing at once – a resonant lived experience of platform navigation, a distributed accomplishment, and a messy data source.
From Navigation to Metadating
What becomes of AI-generated content once it enters the social media platforms for which it was designed and through which it gains traction? It turns into slop not by virtue of its internal synthetic qualities, but through its entanglement with the platform’s environment of expected use. Rather than asking whether a given image – say, Trump as the new Pope – is inherently fake or biased, we can instead examine how countless similar images become operative through the interplay of user appropriation and algorithmic amplification.
Obviously, AI slop generates engagement – views, likes, clicks, and comments. It also establishes connections between contents, exposure time, and recommendations. Crucially, the experience of encountering these recommendations is itself folded back into the cycle of slopification as data that shapes future pathways through algorithmic space. The whole arrangement can be studied through qualitative walkthrough methods that help account for the patterns emerging between user activities and FYP recommendations. Researchers can, for example, create different persona accounts and systematically document the feed with screenshots, tracing how different use scenarios gradually alter the kinds of slop that surface through successive interactions. As Stefanie Duguay and Hannah Gold-Apel suggest, a research persona can serve as a means of accessing specific content flows. A sloppy research persona drives this logic ad absurdum, by promiscuously following, rather than narrowing, recommendation pathways to access a wider spectrum of emergent synthetic trends.
Alternatively, researchers can also ‘listen’ to recommendations. Sal Hagen, Daniël de Zeeuw, and Tommaso Venturini propose digital rhythmanalysis as one such approach. Reflecting on “the evanescence of audio-visual data”, they draw on Lefebvre to conceptualize platform engagement as shaped by “the rhythmic acts of swiping, scrolling, and streaming”. The data that become analytically relevant are generated by repeated encounters with similar-but-different content. To respond to an increasingly AI-driven, ‘post-viral’ Web, rhythmanalysis prioritizes time over space by moving beyond network-centric understandings of digital culture. Sloppy as she is, my research persona commits to this proposition only in part. By dividing attention between rhythmic patterns and networked circulation, she attends to the feel of her “For You” feeds as something inevitably shared, albeit with variations, and thus generative of metadata that defies unifying readings.
From her navigation experience as the main point of departure, a sloppy research persona moves to metadating, though she cannot do so on her own. While inhabiting elp-2026 on my mobile device, I ‘pre-cook’ For You pages of TikTok, YouTube, Instagram, and X for later browser-based data collection using digital methods research tools. Zeeschuimer, developed by Stijn Peeters and colleagues, “to look over your shoulder and collect post metadata as you browse”, serves as an interface for capturing my persona’s recommendations as dynamic data assemblies. Metadata produced through user activity moves across devices, furnishing platforms with the signals needed to organize future circuits of engagement. Watch time wasted on an AI-generated pro-AfD TikTok in the app soon registers in the browser as a recommendation environment populated by similar content. In Metadata, Mon Amour, Lev Manovich explains that metadata allows computers to connect “data with other data.” Metadating is an invitation for researchers to spend time with the multifaceted texture of these connections and engage in practices of ethical and contextual data remix.
What does it mean? Platform metadata attached to algorithmic recommendations – video thumbnails and captions, timestamps and engagement metrics, account names, co-hashtags, comment sections, and sound titles – can be read both quantitatively, as indicators of temporal irregularities and intensities, and qualitatively, as records of affective immersion in platform flows. The very same data are also spatially embedded and relational. Any recommended video that captures attention prompts people to click sound and account buttons, swipe through memetic collections, and share their own contributions with associated keywords. In the perpetually updated “For you” feeds, the algorithmic real-time meets possibilities for prolongation along various networked paths. Comments, for example, are more than a numeric indicator of visibility: they prolong the resonance of recommended content through vernacular engagement with and across singular posts. To illustrate how slopification gives rise to shared audiences, let me return to the U.S. Army toddler-daddy reunion.
The recording shows only a small fraction of the 138 same-themed YouTube Shorts that came to dominate the feed during a 15-minute Zeeschuimer session with elp-2026, in which my research persona was exposed to 560 videos in total. Viewed consecutively, the downloaded videos reveal a shift from content to atmosphere that extends beyond the AI-generated videos themselves into the broader environment of YouTube recommendations and participation. What registered in the browser as a series of fleeting encounters with sappy patriotism emerges as a more durable affective formation sustained by overlapping commenting publics. To trace these relations, I used the co-commenting module of Bernhard Rieder’s YouTube Data Tools on the full video corpus. Analytically, YouTube Shorts comments can be repurposed as edges linking videos. The resulting network maps shared attention, allowing similarities among videos to emerge from audience activity. The network fragment below displays only videos connected by at least five co-commenting accounts, revealing how a distinctive blend of military devotion, family values, and national symbolism travels across associated video milieus.
The analysis is based on a simple premise: if the same user comments on two videos, a connection can be established between them. The more users co-comment across both videos, the stronger the tie. Colors indicate channel affiliations. At the center of the network, most videos appear in grey, indicating they were published by different channels. From here, a dispersed ‘military daddy’ ecology emerges saturated with synthetic adaptations of the same affective script. Made-for-the-moment channels such as Heroes Never Forgotten, Tiny Solgiers, and Patriot Pulse appear alongside kid influencer content from Bablet Official and Bebo Tales. Less a centralized propaganda effort than a diffuse form of ambient amplification, the network illustrates how otherwise unrelated attempts to capture attention feed off the heightened affective intensity surrounding the escalating U.S.-Iran conflict.
In the comment sections, the edges that constitute the network translate into recurrent affective formulas. Across videos, occasional acknowledgments of the scenes’ artificiality do little to disrupt the main tenor of responses: praying for soldiers, blessing families, posting strings of hearts, crying emojis, and declarations of support. Debate over whether the videos are “real” or “fake” rarely gains traction. Instead, emojified shoutouts reappear with ritual regularity, attaching themselves to synthetic family reunions where affect accumulates through phatic bonding. In Updating to Remain the Same, Wendy Hui Kyong Chun insightfully remarks that networks “for all their preoccupation with singular events rely on continual repetitions – or the possibilities of repetition.” Each edge is a “potential interaction based on repeated past interactions”. Slopification is the most pervasive expression of this logic: In the creation of affective attachments, the message matters less than its recurrence. As algorithmic amplification and affective uptake fold into one another, meme wars move from riffing on content to riffing on vibes.
Slopification Mon Amour
“My slop vibes. Look at this For You Page, you’ve got to love how bad it is.” – from a conversation between two students during the same research seminar.
AI slop is a shared object of attachment whose significance lies in its capacity to reenact familiar affective orientations – on repeat. This brings us to riffing as the driving force of slopification. Rather than mere pattern variation in content formations, riffing plays out between user practices and algorithmic systems. The term comes from sound studies, where a “riff” is a repeated musical phrase, still recognizable to sustain familiarity, but different enough to capture attention. Theorists such as Steve Goodman have linked riffing to the "bad vibes" of memetic warfare, while Robin James’s account of “good vibes” suggests that algorithms increasingly take on a similar operational trajectory, amplifying emergent orientations through vectors and probabilities.
As a form of the “changing same,” riffing turns such affective in-betweenness into a source of value production. We do not love or hate AI slop because it is intrinsically good or bad, but because it places us in an affective condition where attachment persists even as the objects that sustain it become increasingly disposable. Riffing is a reciprocal process, and if, as Sara Ahmed argues, love is an investment whose value lies in the promise of return, then, maybe, contemporary anxieties around slop reveal our overstimulated attachment to the internet itself. We remain attached to platforms as sites of affective bonding, while the proliferation of synthetic content increasingly appears as evidence that this investment is being reciprocated at scale. The internet delivers the promised object, but in overwhelming abundance. It’s just too much. We receive endless variations of the same affective formula, reproduced faster than ever before.
This helps explain the peculiar ambivalence surrounding AI slop. On the one hand, it is frequently experienced as a degradation of online culture: repetitive, manipulative, algorithmically flattened, and fake, fake, fake. On the other hand, it demonstrates an extraordinary capacity to organize attention. The military daddy ecology, though by no means exceptional, illustrates this dynamic particularly well. Synthetic scenes of family reunion, sacrifice, and patriotism circulate across channels and audiences, accumulating responses despite the obvious artificiality of the object of interest. The synthetic soldier returning home provides a low-cost, infinitely reproducible surface onto which conflicting affective investments – war anxieties and love for the nation – can be casually projected by anyone or anything capable of engagement. The accounts involved could be bots or real people. Synthetic toddlers could just as easily be cute cats or puppies, but that hardly matters. Slop annoys and captivates. Slop distracts. Slop succeeds because, as a distributed accomplishment, it effortlessly blurs *so many feels* in anticipation of the next sloppy encounter.
--
This research builds on a talk for the lecture series “Cooperative -Methodologies – Studying Sensory Media and AI” at the DFG-funded Research Center “Media of Cooperation”. It is a result of collaborative teaching. Thanks to the students and to Sergei Pashakhin for his wit and for adding a YouTube Shorts fork to Zeeschuimer.
Bio
Elena Pilipets is a Digital Media and Methods research fellow at the University of Siegen with a PhD in Media Studies from the University of Klagenfurt. Her work integrates media theory with qualitative, digital, and AI-based methods for studying contemporary platform cultures.
References
Aris Komporozos-Athanasiou, Speculative Communities: Living with Uncertainty in a Financialized World, Chicago: The University of Chicago Press, 2022.
Arte Tracks, The Death of the Internet, YouTube video, 2025, https://www.youtube.com/watch?v=Z5LBwoCNwQo
Bernhard Rieder, Engines of Order: A Mechanology of Algorithmic Techniques, Amsterdam University Press, 2020.
Bernhard Rieder, YouTube Data Tools (Version 1.43) [Software], 2015, https://ytdt.digitalmethods.net
Carl T. Bergstrom and Jevin D. West, ‘Modern-day oracles or bullshit machines? How to Thrive in a ChatGPT World, online course, The Bullshit Machines, https://thebullshitmachines.com
Chris Nodder, Evil by Design: Interaction Design to Lead Us into Temptation, Hoboken, NJ: Wiley, 2013.
Clemens Apprich, Wendy Hui Kyong Chun, Florian Cramer, and Hito Steyerl, Pattern Discrimination, Minneapolis: University of Minnesota Press, 2018.
Dan Brooks, ‘The Internet’s Newest Hate Watch’, The New York Times, 1 June 2026, https://www.nytimes.com/2026/06/01/opinion/fruit-love-island-ai-slop.html
Dawn Lyon and Rebecca Coleman, ‘Rupture, Repetition, and New Rhythms for Pandemic Times: Mass Observation, Everyday Life, and COVID-19’, History of the Human Sciences 36.2 (2023): 3–24.
Elena Pilipets, ‘Hashtagging, Duetting, Sound-linking: TikTok Gestures and Methods of (In)distinction’, MAST: The Journal of Media Art Study and Theory 4.1 (2023): 109–135.
Elena Pilipets, Sofia P. Caldeira, and Ana Marta M. Flores, ‘“24/7 horny&training”: Porn Bots, Authenticity, and Social Automation on Instagram’, Porn Studies (2024), https://doi.org/10.1080/23268743.2024.2362166
Elena Pilipets, Marloes Geboers, ‘Ambient Amplification: Memecry Across Platforms.’ Platforms & Society (2026 forthcoming)
Eryk Salvaggio, ‘How to Read an AI Image: The Datafication of a Kiss’, Cybernetic Forests, 2 October 2022, https://cyberneticforests.substack.com/p/how-to-read-an-ai-image
Gregory Chatonsky, ‘The End of Memes’, Reflection, July 2025, https://chatonsky.net/la-fin-des-memes-the-end-of-memes/
Gregory Chatonsky, ‘Vectofascism – Part I: Vectoaesthetics’, Art in the Age of Average: The New AI-thoritarians, Carrier-Bag, 4 February 2026, https://carrier-bag.net/vectofascism-part-1-vectoaesthetics/
Holly Herndon and Mat Dryhurst, All Media is Training Data, edited by Eva Jäger and Caroline Busta. London: Serpentine, 2024.
Jason Koebler, ‘AI Slop Is a Brute Force Attack on the Algorithms That Control Reality’, 404 Media, 17 March 2025, https://www.404media.co/ai-slop-is-a-brute-force-attack-on-the-algorithms-that-control-reality/
James Burton et al., Algorithmic Authenticity, Lüneburg: Meson Press, 2023.
Joan Donovan, Emily Dreyfuss, and Brian Friedberg, Meme Wars: The Untold Story of the Online Battles Upending Democracy in America, Bloomsbury Publishing, 2022.
Jodi Dean, ‘Faces as Commons: The Secondary Visuality of Communicative Capitalism’, Open!, 31 December 2016, https://onlineopen.org/faces-as-commons
Kate Crawford, ‘Metabolic Images’, Aperture Archive, Winter 2024, https://archive.aperture.org/article/2024/4/4/kate-crawford-metabolic-images
Lev Manovich, ‘Metadata, Mon Amour’, 2002, https://manovich.net/index.php/projects/metadata-mon-amour
Liliana Bounegru, Melody Devries, and Esther Weltevrede, ‘The Research Persona Method: Figuring and Reconfiguring Personalised Information Flows’, in Celia Lury, William Viney, and Scott Wark (eds), Figure: Concept and Method Singapore: Palgrave Macmillan, 2022, pp. 77–104.
Ludmila Lupinacci, ‘Phenomenal Algorhythms: The Sensorial Orchestration of “Real-Time” in the Social Media Manifold’, New Media & Society 26.7 (2024): 4075–4093.
Marcus Bösch and Tom Divon, ‘Dancing in Uniform: Thirst Trap Propaganda in Military Image Wars on TikTok’, Platforms & Society (2026), https://doi.org/10.1177/29768624251403864
Mike Isaac and Eli Tan, ‘OpenAI’s New Video App Is Jaw-Dropping (for Better and Worse)’, The New York Times, 2 October 2025, https://www.nytimes.com/2025/10/02/technology/openai-video-app.html
Minna Ruckenstein, The Feel of Algorithms, University of California Press, 2023.
Mitch Therieau, ‘The Novel of Vibes’, Post45, March 2022, https://post45.org/2022/03/the-novel-of-vibes/#footnote_5_17212
Nick Seaver, ‘Everything Lies in a Space: Cultural Data and Spatial Reality’, Journal of the Royal Anthropological Institute 27.1 (2021): 43–61.
Olga Goriunova, ‘The Lurker and the Politics of Knowledge in Data Culture’, International Journal of Communication 11 (2017): 3917–3933.
Oliver Whang, ‘Can You Get Rich Quick Off A.I. Slop?’, The New York Times Magazine, 1 June 2026, https://www.nytimes.com/interactive/2026/06/01/magazine/ai-slop-viral-videos.html
Paul Bouchaud, Grok Unleashed, AI Forensics Report, 2025, https://aiforensics.org/uploads/GrokUnleashed.pdf
Paul Roquet, ‘In the Mood: Toward the Full Commodification of Ambience’, Real Life, 26 April 2021, https://reallifemag.com/in-the-mood/
Petter Törnberg and Richard Rogers, ‘Towards a Post-Social Media Studies’, OSF Preprints, 21 April 2026, https://osf.io/preprints/socarxiv/6nue7_v1
Robin James, Good Vibes Only: Phenomenology and the Biopolitics of Algorithmic Legitimation, Durham, NC: Duke University Press, 2026.
Roland Meyer, AI Slop and the New Fascist Aesthetic, Episode 62, Open Spotify, 11 March 2025, https://open.spotify.com/episode/5XxfdDFetoiYRWqkAdtW3w
Sal Hagen, Daniël de Zeeuw, and Tommaso Venturini, ‘Digital Rhythmanalysis: Studying Memetic and Affective Rhythms on the Post-Viral Web’, Platforms & Society (2025), https://doi.org/10.1177/29768624251394
Sara Ahmed, The Cultural Politics of Emotion, Edinburgh: Edinburgh University Press, 2004.
Stefanie Duguay and Hannah Gold-Apel, ‘Stumbling Blocks and Alternative Paths: Reconsidering the Walkthrough Method for Analyzing Apps’, Social Media + Society 9.1 (2023), https://doi.org/10.1177/20563051231158822
Steve Goodman, Sonic Warfare: Sound, Affect, and the Ecology of Fear, Cambridge, MA: MIT Press, 2009.
Stijn Peeters, Zeeschuimer [browser extension], 2022, https://github.com/digitalmethodsinitiative/zeeschuimer
Susanna Paasonen, ‘Fickle Focus: Distraction, Affect and the Production of Value in Social Media’, First Monday 21.10 (2016), https://doi.org/10.5210/fm.v21i10.6949
Susanna Paasonen, ‘Ambiguous Affect: Excitements That Make the Self’, in Gregory J. Seigworth and Carolyn Pedwell (eds.), The Affect Theory Reader 2: Worldings,
Tensions, Futures, Durham, NC: Duke University Press, 2023, pp. 85–102.
Taina Bucher, If… then: Algorithmic Power and Politics, Oxford: Oxford University Press, 2018.
Wendy Hui Kyong Chun, Updating to Remain the Same: Habitual New Media, MIT Press, 2016.