Bot farms invade social media to hijack popular sentiment

Welcome to the world of social media mind control. By amplifying free speech with fake speech, you can numb the brain into believing just about anything. Surrender your blissful ignorance and swallow the red pill. You’re about to discover how your thinking is being engineered by modern masters of deception.

The means by which information gets drilled into our psyches has become automated. Lies are yesterday’s problem. Today’s problem is the use of bot farms to trick social media algorithms into making people believe those lies are true. A lie repeated often enough becomes truth.

“We know that China, Iran, Russia, Turkey, and North Korea are using bot networks to amplify narratives all over the world,” says Ran Farhi, CEO of Xpoz, a threat detection platform that uncovers coordinated attempts to spread lies and manipulate public opinion in politics and beyond.

Bot farm amplification is being used to make ideas on social media seem more popular than they really are. A bot farm consists of hundreds and thousands of smartphones controlled by one computer. In data-center-like facilities, racks of phones use fake social media accounts and mobile apps to share and engage. The bot farm broadcasts coordinated likes, comments, and shares to make it seem as if a lot of people are excited or upset about something like a volatile stock, a global travesty, or celebrity gossip—even though they’re not.

Meta calls it “coordinated inauthentic behavior.” It fools the social network’s algorithm into showing the post to more people because the system thinks it’s trending. Since the fake accounts pass the Turing test, they escape detection.

Unlike conventional bots that rely on software and API access, bot farm amplification uses real mobile phones. Racks and racks of smartphones connected to USB hubs are set up with SIM cards, mobile proxies, IP geolocation spoofing, and device fingerprinting. That makes them far more difficult to detect than the bots of yore.

“It’s very difficult to distinguish between authentic activity and inauthentic activity,” says Adam Sohn, CEO of Narravance, a social media threat intelligence firm with major social networks as clients. “It’s hard for us, and we’re one of the best at it in the world.”

Disinformation, Depression-era-style

Distorting public perception is hardly a new phenomenon. But in the old days, it was a highly manual process. Just months before the 1929 stock market crash, Joseph P. Kennedy, JFK’s father, got richer by manipulating the capital markets. He was part of a secret trading pool of wealthy investors who used coordinated buying and media hype to artificially pump the price of Radio Corp. of America shares to astronomical levels.

After that, Kennedy and his rich friends dumped their RCA shares at a huge profit, the stock collapsed, and everyone else lost their asses. After the market crashed, President Franklin D. Roosevelt made Kennedy the first chairman of the Securities and Exchange Commission, putting the fox in charge of the henhouse.

Today, stock market manipulators use bot farms to amplify fake posts about “hot” stocks on Reddit, Discord, and X. Bot networks target messages laced with ticker symbols and codified slang phrases like “c’mon fam,” “buy the dip,” “load up now” and “keep pushing.” The self-proclaimed finfluencers behind the schemes are making millions in profit by coordinating armies of avatars, sock puppets, and bots to hype thinly traded stocks so they can scalp a vig after the price increases.

“We find so many instances where there’s no news story,” says Adam Wasserman, CFO of Narravance. “There’s no technical indicator. There are just bots posting things like ‘this stock’s going to the moon’ and ‘greatest stock, pulling out of my 401k.’ But they aren’t real people. It’s all fake.”

In a world where all information is now suspect and decisions are based on sentiment, bot farm amplification has democratized market manipulation. But stock trading is only one application. Anyone can use bot farms to influence how we invest, make purchasing decisions, or vote.

These are the same strategies behind propaganda efforts pioneered by Russia and the Islamic State to broadcast beheadings and sway elections. But they’ve been honed to sell stocks, incite riots, and, allegedly, even tarnish celebrity reputations. It’s the same song, just different lyrics.

“People under the age of 30 don’t go to Google anymore,” says Jacki Alexander, CEO at pro-Israel media watchdog HonestReporting. “They go to TikTok and Instagram and search for the question they want to answer. It requires zero critical thinking skills but somehow feels more authentic. You feel like you’re making your own decisions by deciding which videos to watch, but you’re actually being fed propaganda that’s been created to skew your point of view.”

It’s not an easy problem to solve. And social media companies appear to be buckling to inauthenticity. After Elon Musk took over Twitter (now X), the company fired much of its anti-misinformation team and reduced the transparency of platform manipulation. Meta is phasing out third-party fact-checking. And YouTube rolled back features meant to combat misinformation.

“On TikTok, you used to be able to see how many times a specific hashtag was shared or commented on,” says Alexander. “But they took those numbers down after news articles came out showing the exact same pro-Chinese propaganda videos got pushed up way more on TikTok than Instagram.” Algorithmically boosting specific content is a practice known as “heating.”

If there’s no trustworthy information, what we think will likely become less important than how we feel. That’s why we’re regressing from the Age of Science—when critical thinking and evidence-based reasoning were central—back to something resembling the Edwardian era, which was driven more by emotional reasoning and deference to authority.

When Twitter introduced microblogging, it was liberating. We all thought it was a knowledge amplifier. We watched it fuel a pro-democracy movement that swept across the Middle East and North Africa called the Arab Spring and stoke national outrage over racial injustice in Ferguson, Missouri, planting the seeds for Black Lives Matter.

While Twitter founders Evan Williams and Jack Dorsey thought they were building a platform for political and social activism, their trust and safety team was getting overwhelmed with abuse. “It’s like they never really read Lord of the Flies. People who don’t study literature or history, they don’t have any idea of what could happen,” said tech journalist Kara Swisher in Breaking the Bird, a CNN documentary about Twitter.

Your outrage has been cultivated

Whatever gets the most likes, comments, and shares gets amplified. Emotionally charged posts that lure the most engagement get pushed up to the top of the news feed. Enrage to engage is a strategy. “Social media manipulation has become very sophisticated,” says Wendy Sachs, director-producer of October 8, a documentary about the campus protests that erupted the day after the October 7th Hamas attack on Israel. “It’s paid for and funded by foreign governments looking to divide the American people.”

Malicious actors engineer virality by establishing bots that leach inside communities for months, sometimes years, before they get activated. The bots are given profile pics and bios. Other tricks include staggering bot activity to occur in local time zones, using U.S. device fingerprinting techniques like setting the smartphone’s internal clock to the time zone to where an imaginary “user” supposedly lives, and setting the phone’s language to English.

Using AI-driven personas with interests like cryptocurrency or dogs, bots are set to follow real Americans and cross-engage with other bots to build up perceived credibility. It’s a concept known as social graph engineering, which involves infiltrating broad interest communities that align with certain biases, such as left- or right-leaning politics.

For example, the K-pop network BTS Army has been mobilized by liberals, while Formula 1 networks might lend themselves to similar manipulation from conservatives. The bots leach inside these communities and earn trust through participation before agitating from within.

“Bot accounts lay dormant, and at a certain point, they wake up and start to post synchronously, which is what we’ve observed they actually do,” says Valentin Châtelet, research associate at the Digital Forensic Research Lab of the Atlantic Council. “They like the same post to increase its engagement artificially.”

To set up and manage thousands of fake social media accounts, some analysts believe that bot farms are deployed with the help of cyber criminal syndicates and state sponsorship, giving them access to means that are not ordinary.

View this post on Instagram

A post shared by Jack Latham (@jacklatham)

Bot handlers build workflows with enough randomness to make them seem organic. They set them to randomly reshare or comment on trending posts with certain keywords or hashtags, which the algorithm then uses to personalize the bot’s home feed with similar posts. The bot can then comment on home feed posts, stay on topic, and dwell deeper inside the community.

This workflow is repetitive, but the constant updates on the social media platform make the bot activity look organic. Since social media platforms update frequently, programmed bots appear spontaneous and natural.

Software bots posted spam, also known as copypasta, which is a block of text that gets repeatedly copied and pasted. But the bot farmers use AI to author unique, personalized posts and comments. Integrating platforms like ChatGPT, Gemini, and Claude into a visual flow-building platform like Make.com, bots can be programmed with advanced logic and conditional paths and use deep integrations leveraging large language models to sound like a 35-year-old libertarian schoolteacher from the Northwest, or a MAGA auto mechanic from the Dakotas.

The speed at which AI image creation is developing dramatically outpaces the speed at which social networking algorithms are advancing. “Social media algorithms are not evolving quick enough to outperform bots and AI,” says Pratik Ratadiya, a researcher with two advanced degrees in computer science who’s worked at JPL and Apple, and who currently leads machine learning at Narravance. “So you have a bunch of accounts, influencers, and state actors who easily know how to game the system. In the game of cat and mouse, the mice are winning.”

Commercial bot farms charge around a penny per action to like posts, follow accounts, leave comments, watch videos, and visit websites. Sometimes positioned as “growth marketing services” or “social media marketing panels,” commercial bot farms ply their trades on gig marketplaces like Fiverr and Upwork, as well as direct to consumer through their own websites.

Bot farms thrive in countries where phone numbers, electricity, and labor are cheap, and government oversight is minimal. A new book about Vietnamese bot farms by photographer Jack Latham called Beggar’s Honey offers a rare look into Hanoi’s attention sweatshops.

Sentiment breeding

Before free speech colluded with fake speech online, traders used fundamental analysis and technical analysis to decide what and when to buy. Today, fund managers and quants also use sentiment analysis to gauge the market’s mood and anticipate volatility. Sentiment scores are even fed into trading algorithms to automatically adjust exposure and trigger trades.

Courts recently