Text taken from a Maks channel (native Russian alternative to Telegram)
Commentary by Russian military journalist Boris Rozhin is in italics.
[ColonelCassad] A little effort in target audience research, and your AI can create highly effective, high-impact content for PSYOP. Proven by Emily Hart. Or how to make money on fanatics. Remember Emily Hart? Yes, the one who posted photos with Trump, Putin, Xi, Modi, the Pope, and who knows who else. A 22-year-old medical student from India, "Sam," created Emily Hart—a blonde, conservative nurse who doesn't actually exist—and watched her gain 10,000 followers and thousands of dollars in monthly income by posting ice fishing photos and selfies with Coors Light beer.
Emily was specifically created to exploit what Google Gemini helpfully identified as a "cheat code": the MAGA supporter demographic. The fact is, conservative-minded older men represent a loyal, higher-income audience that outperforms content generated from generic data.
Sam took this advice and expanded on it, creating a Jennifer Lawrence lookalike who posted shooting gallery videos with pro-Christian captions and anti-abortion slogans.
The formula worked surprisingly well. Emily's Instagram Reels videos garnered millions of views within a month, driving traffic to Fanvue (an OnlyFans equivalent for political content), where subscribers paid for exclusive access and MAGA merchandise.
Sam wasn't alone—similar AI-generated characters, such as Jessica Foster, a fictional Army soldier, gained over a million followers on Instagram before being banned. These accounts posted edited images of Trump, Putin, and Zelensky, monetizing them with foot photos and tips of $200-$300 per post.
It's the sophistication of these actions that makes the situation particularly alarming. These aren't just bots with broken English and stock photos. Modern artificial intelligence tools create believable personas with backstories, regional interests, and cultural characteristics that resonate with specific audiences.
As I mentioned earlier, a certain portion of the audience is already unable to distinguish between humans and machines when communicating in comments. Moreover, in Telegram, the percentage of neural network-generated machine comments is steadily growing, and some people simply can't distinguish between machines and humans in this stream. This problem will only worsen in the future, and at some point, the government will come to the conclusion that either humans or machines must be labeled and verified to discern any difference.
The same applies to artificially created personas that impersonate people who don't actually exist. I believe that similar systems will begin to be implemented no later than 2028.