Every election has its own kind of morning-after feeling — a mix of relief, exhaustion, and hope that the shouting is over and the counting is done. But after the recent midterm elections, something else lingered beneath the surface: a quiet, coordinated attempt to make people doubt the results before the results were even final. According to a new report from Cyabra, an artificial intelligence company cited by the Washington Examiner, thousands of fake social media accounts were activated around Election Day to spread fabricated vote totals and misleading claims about key races. This wasn’t necessarily an attempt to change who won or lost. It was something more insidious — an attempt to poison the well of public trust itself. The report found that nearly one in five Election Day posts on X, the platform formerly known as Twitter, about five closely watched races came from inauthentic accounts. That’s a startling number. It means that for every five posts a voter saw about those races, one was likely not from a real person. And because the fake accounts were designed to look ordinary and blend in, most people probably had no idea they were being manipulated. The goal, as the report puts it, was not any single race but trust in the result itself. In a democracy, that distinction matters more than most people realize.
The numbers behind the report are worth sitting with for a moment. Cyabra identified 8,344 suspicious accounts that were active between October 20 and November 3. Of those, 84 percent were classified as “sleeper” accounts — profiles that had not shown up in earlier monitoring and appeared to be waiting quietly until the right moment to spring into action. What makes this especially concerning is that most of these sleeper accounts were created before October, not in a sudden burst of last-minute registrations. Only 76 of them were created after October 1. That suggests a deliberate strategy: set up the accounts well in advance, let them age and look normal, then activate them as Election Day approaches. This is the modern evolution of political manipulation. In the past, you might have seen obvious bots with strange usernames and no profile pictures. Now, the fake accounts look like your neighbors, your coworkers, your old classmates. They share memes, complain about the weather, post about sports, and then, on the one day that matters most, they start spreading false vote totals and sowing confusion. The report also found that nearly 85 percent of the X accounts with identifiable locations were based in the United States. That doesn’t rule out foreign involvement entirely, but it does suggest that a significant portion of this activity may be homegrown. It’s a reminder that disinformation isn’t always a foreign boogeyman — sometimes it’s just people or organizations willing to sacrifice democratic stability for political advantage.
The five races targeted by this suspicious activity were Senate contests in Michigan, Maine, and Kansas, along with gubernatorial races in Ohio and Georgia. These are all states where the outcomes mattered deeply, not just to the candidates but to the national political landscape. The report found that both parties were hit, which suggests that the operation was less about supporting a particular candidate and more about generating a general sense that the system couldn’t be trusted. When both sides are victims, the real beneficiary is chaos. The activity was heavily concentrated on X, with substantially less detected on Facebook and Instagram. TikTok was excluded from the investigation entirely, which means the full scope of the problem could be even larger than what Cyabra was able to document. That uneven distribution is worth noting. It could reflect differences in how platforms monitor and remove fake accounts, or it could reflect where the bad actors believe they can have the most influence. Either way, it’s a reminder that social media platforms are not neutral spaces. They are environments shaped by algorithms, policies, and enforcement decisions. And when one platform becomes the primary vehicle for disinformation, it creates a serious challenge for anyone trying to help voters make sense of what they’re seeing.
Perhaps the most disturbing tactic described in the report involved the use of fabricated vote totals posted before the polls even closed. In Michigan, Cyabra found a fake account that posted bogus Senate results at 1:35 p.m. on Election Day — hours before any legitimate counting could possibly be complete. To make the lie seem more credible, the account asked the AI chatbot Grok to “confirm this.” The company warned that invoking AI in this way can give false claims an appearance of legitimacy, regardless of what the chatbot actually says in response. People see a name like Grok or ChatGPT and assume there is some underlying intelligence verifying the information. But these chatbots are not election officials. They don’t have access to real-time vote totals. They are simply generating responses based on patterns in the data they’ve been trained on. That makes them easy tools for bad actors who want to launder misinformation through a veneer of technological authority. The same fabricated results that appeared at 1:35 p.m. later showed up on a legitimate account at 6:25 p.m., still before the polls closed. Similar activity was detected in Ohio. The pattern is clear: create a false story, get people talking about it, then let it spread organically through confused and concerned citizens who genuinely believe they’re sharing breaking news.
What does all of this mean for the future? Cyabra warns that the tactics are likely to become more sophisticated, especially as the 2028 presidential election approaches. The company expects “the same playbook, faster and closer to the count: fresh sleeper accounts each cycle, AI chatbots invoked as a credibility stamp, and the same accounts working many races at once.” That’s a sobering forecast. It suggests that the disinformation campaigns we saw in the midterms were not an isolated incident but a training ground for something bigger. The people behind these operations are learning. They are testing what works, measuring how quickly false claims spread, and refining their methods. And they are doing it because they know that a democracy doesn’t collapse in a single moment. It collapses slowly, through a thousand small fractures in public confidence. When voters begin to believe that elections are rigged, that their neighbors can’t be trusted, that the counting process is corrupt, they stop participating. They stop believing that their voice matters. And that is exactly what the bad actors want. The report found no evidence that the suspicious activity changed the outcome of any race, which is good news. But changing the outcome may never have been the point. The point was to create doubt, to make people wonder, to plant a seed of suspicion that could grow long after the election was over. And in that sense, the campaign may have already succeeded in ways that are difficult to measure.
So what can be done? Cyabra urges election officials to take online disinformation seriously and to remember that official results come from election authorities, not from social media platforms or AI chatbots. That sounds simple, but it’s actually a profound challenge in a world where many people get their news from their feeds and trust what they see there more than what they hear from institutions. The report’s recommendation is not just about posting press releases or holding press conferences. It’s about building a more resilient information ecosystem. That means working with platforms to identify and remove inauthentic accounts faster. It means creating clear, accessible channels where voters can verify official results in real time. It means educating the public about how to spot fake accounts and why they should be skeptical of vote totals that appear before polls close. But ultimately, the responsibility also falls on all of us as citizens. We need to pause before sharing something that makes us angry or afraid. We need to ask ourselves whether the account that posted it seems real. We need to remember that democracy is not just a system of laws and elections; it is a shared trust. That trust can be broken by a single well-timed lie, but it can also be rebuilt through transparency, patience, and a commitment to the truth. The midterms were a warning. The question is whether we are willing to listen.