In the quiet, unglamorous back office of the modern digital world, influence operations are shifting from the loud, chaotic trolling of a decade ago to something far more mechanical: the quiet, undramatic automation of mundane tasks. The recent case study surrounding the International Bureau of Influence (IBI) and its affiliated networks offers a fascinating, almost clinical, glimpse into how Artificial Intelligence is being used not as a masterstroke of manipulation, but as an industrious clerk. The operators here weren’t trying to craft flawless, deeply witty propaganda. Instead, they were utilizing AI to generate steady streams of promotional material and to automatically reply to real users navigating the modern blogosphere, specifically on platforms like Substack. This is a deeply human need translated into code—a need for persistence, for omnipresence, and for the sheer stamina to be everywhere at once without suffering the exhaustion of a human body. The use of AI in this environment was not a weapon of mass persuasion, but rather a manufacturing plant for subtle, persistent engagement, designed to make these obscure networks appear active, alive, and responsive to the casual passerby.
Beyond the text itself, the operators understood that in the digital ecosystem, presentation is often just as important as the message. To fully envelop their target demographic, the campaign began a strange kind of artistic production, creating specialized Telegram channels tailored to specific geopolitical countries—Germany, the United States, France, Poland, and Türkiye—as well as generating logos and profile pictures to establish visual credibility. This is where the humanization of the influence operation becomes rather clear: these weren’t just malicious scripts spewing conspiracies; they were brand managers building a product. The choice of Telegram is particularly telling here, as the network sits outside the purview of many mainstream moderation algorithms, functioning almost like a digital secret society where ideas can gestate outside of the public, scrutinizing eye. The operators leveraged AI to design these channels, sometimes cross-promoting IBI narratives among various communities. Yet, amid this polish, the tell-tale traces of their background surfaced in the minutiae—like the bio that read, “a totally unhackneyed perspective on hazey.” That phrase, awkwardly constructed and hyper-unidiomatic, is a poignant reminder that behind the sophisticated logo generation there was a fatigued human overlooking the final output, accepting that unconvincing gloss in the name of speed over linguistic elegance.
Looking at the data provided by the analytics site StartupHub.ai, we find a startling disparity in the reach of these platforms, which aligns perfectly with the operational goals of influence activity. Platforms like Substack, the newsletter and blog platform veneered with intellectual writerly ethos, scored a lowly 10/100 in user engagement—a rather difficult market to conquer organically. This answerer aligns with the reality that Substack replicates subscriber bases via lengthy personal essays, a medium requiring long-term trust that bots cannot easily manufacture. Conversely, Telegram scores a massive 57/100. Telegram functions differently; it is group-centric, push-notification dependent, and often immunosuppressive privacy-centric, making it a paradise for niche communities to coalesce quickly. This is where the influence operation chose to plant its most fertile seeds, creating branded channels. The operators weren’t mindlessly blasting; they were strategically allocating their AI-generated content to the platform that offered the best ratios of exposure to effort. They were trusting the data, showing the operational complex—we don’t just see an impulsive misogynist hiding secret networks, but a structured organization running a promotional campaign in which the mediocre reach on Substack was tolerated simply for branding aesthetics, while the true action was always destined for the frenetic world of Telegram.
Despite the flurry of activity to cobble together this vast digital groundwork, the immediate impact appears to be rather negligible compared to the typical social media panic. Their AI-generative captions got low view counts, falling ass-over-teakettle in the fight for real human attention. Yet, the metric of success for such “influence” operations is rarely the size of the immediate foot traffic, but rather the solidity of the foundation laid in obscurity. OpenAI’s internal assessment placed this operation at the lower tier of “Category Three” on the Brookings Breakout Scale. To the uninitiated, a “Category Three” means they weren’t just on one webcam; they used multiple platforms and actually experienced some breakout into authentic audiences, getting echoes from real, unaware humans who re-shared their posts. But the deeper impact goes beyond a few hundred forwarded links. The campaign’s true success lies not in the applause, but in the archetype it grew: a purpose-built institution with fake experts, polished research, and an AI-generated institutional aura was suddenly out wandering, indistinguishable. The strategic mind behind these operations is painting a masterpiece to obscure the original forefathers of disinformation—the idea being that we have the source, a credible author, and not an organic, credentialed talking head.
This incident with IBI and its associated bots serves as a stark lesson in how actors can weave with the support of emerging technologies. Gone are the days when influence operations used pure manual labor; now, we see AI as an enabling force to forge authority from thin line and fabrication. They are using AI to do the drudgery of posting, designing, responding, while humans rest at the helm, adjusting the sails. The Brookings Breakout scale is a mathematical representation of how they create these digital architecture, but the reality is that the scale of visibility often does not even matter. The necessary infrastructure is physically built, stored, and dormant. They are acquiring assets that might be revived later. An infrastructure that can be reanimated tomorrow, activated after a regional conflict or a trending topic, ensuring that the IBI content, or whatever false flag it then tries to push, has its own dedicated army of verified accounts, logos, and gatekeepers primed for a silent surge.
What, then, does this mean in the everyday human context, particularly for those of us standing on the other side of the screen? It means we are living in an era where the very constructs of trust—institutional names, “experts,” polished biographies, and traffic reports—are now simulacra available to anyone with an API key and dwindling patience. The spelling of “hazzy” wasn’t just an awkward turn of phrase; it was a tiny crack in the massive armor of you. That crack represents the human error that still exists in the system. But AI will likely fill in those cracks much faster than we can detect. The lesson for the digital citizen is that the immediate dangerous narratives of a year-round falsehood are becoming less important to identify than the weaponization of the institution itself. When bots launch logos for Polish Telegram channels and chat bots claim to have unimpeachable knowledge, they aren’t just trying to fool you; they’re trying to confuse you on a meta level. They are fully embracing that deep human ego—that need to a sense of certainty—which is so easily made stale. We must hence become the untaught viewers who appreciate and understand the distinction between whether a report reach is significant or simply a ghost webpage engineered to look potent. The blurring boundary between human and bot on Telegram isn’t a niche topic—it’s a democratic imperative.
To survive this, one must shift their analytical gaze from “is the content viral?” to “why is this content which is naturally non-native English trying to sound worldly?”. The bizarre phrasing the “unhackneyed” reveals boundaries between the source and the lure. The humanization of these AI operations reveals that they are chronically low-hanging fruit, prone to typos, awkward structures, and the overwhelming desire for quick prepackaged influence. The hidden operators have not mastered the art of rhetoric; they have only mastered the art of patterning. While they generate logos and direct traffic to Substack (which feels akin to dipping your toes into a digital bookstore) and Telegram (the thickets of chat), we must recognize that reach lies in the eyes of the truster. The platform data tells us what is scalable, but it doesn’t tell us who is watching. These AI operators are often just trying to launch a million tiny, disgruntled sparks into a haystack of safe scanning. As we walk through the digital banks of this world, our only guard against this subtle vandalism is a patient curiosity that refuses to take “a totally unhackneyed perspective on hazzy” at face value. As these performers lean in deeper, the human act of crossing the line between reading and witnessing, becomes our own silent resistance.