China-linked bot farm anti-data-center campaign on X had little real impact, researcher finds

X recently announced it had uncovered a China-linked bot farm of roughly 200,000 accounts, with a subset of about 200 accounts posting anti-data-center content aimed at influencing U.S. AI and energy policy. But independent analysis by Clemson University researcher Darren Linvill tells a different story: at least 59 of those bots had zero engagement and were likely deactivated before anyone saw their posts. The practical takeaway for anyone following AI infrastructure debates is that this operation almost certainly didn't move the needle.

What the data actually shows

Linvill, co-director of Clemson's Media Forensics Hub, used a platform that retains data from suspended X accounts to search for the hashtag "capacityauction" used by the bots. He found 90 posts, all created in January. Each account posted only two or three times before X suspended them. Linvill told Business Insider that "no real humans saw these posts about data centers" and called the operation a "non-story."

The bot accounts were part of a larger farm of 200,000 accounts, but only a couple hundred focused on data centers. X's own disclosure noted the accounts posted content about data centers driving up electricity bills and straining the energy grid. Linvill also noted that some of these posts were the same ones OpenAI flagged in a June report about Chinese accounts using ChatGPT to generate anti-data-center content.

Why the limited reach matters for AI policy debates

The broader context is that genuine grassroots opposition to AI data centers is widespread and bipartisan. A Gallup poll from May found over 70% of respondents oppose data center construction in their area, citing energy and water concerns. Politicians have responded with moratoriums and new regulations. Texas Governor Greg Abbott, once a supporter, has instructed state agencies to protect residents from infrastructure costs, claiming his new requirements prevented 1,800 AI data centers from being built.

University of Pittsburgh researcher Samuel Woolley pointed out that the only reports of Chinese propaganda on this topic have come from Big Tech companies with a "vested interest" in data center development. X is owned by Elon Musk, whose companies rely on data centers. This doesn't mean the bot farm didn't exist, but it does mean the narrative of a foreign influence campaign driving anti-data-center sentiment is not supported by the engagement data.

Practical implications for evaluating influence operations

For AI builders and operators watching policy debates, this case is a reminder to distinguish between claimed coordination and demonstrable reach. A bot farm with 200,000 accounts sounds alarming, but if the subset targeting a specific debate had zero engagement, the actual influence is negligible. When platforms disclose inauthentic behavior, it's worth waiting for independent verification of reach before concluding that public opinion was swayed.

The same principle applies to any claim about social media manipulation in AI policy discussions. Look for evidence of real human engagement, not just account counts.

What remains unclear

The total number of bot accounts and their full reach may evolve as more data becomes available. X's disclosure relied on its own investigation, and independent researchers have only analyzed a portion of the suspended accounts. The connection between this bot farm and Chinese state actors has not been independently confirmed, though both X and OpenAI have attributed the activity to China.

The key caveat: the available evidence suggests this operation had minimal impact on the U.S. data center debate. But that doesn't mean other, more effective influence campaigns don't exist. It just means this particular one was a rowboat trying to move an oil tanker.