Rahil Prakash is only 21, a recent college graduate still finding his way in the world, and he became, for a few strange days, the invisible author of something that looked like a credible political research organization. The website was called Median Strategies, and it published polling that looked polished enough to be real, with names, numbers, percentages, and a tone of neutral authority. Some of those fabricated polls found their way into public conversation and even into the announcements of campaigns. The operation collapsed this week after a reporter from the Los Angeles Times started asking questions, and the site quickly took down its findings. In its final unsigned statement, Median Strategies called the project a “short-term social experiment” intended to see whether fake polling could slip into the political information system without the sort of independent review society expects from journalism. The creator behind it all, when reached by phone, sounded neither triumphant nor entirely ashamed. He said he had wanted to test a theory. “I wanted to see if fake polls could really penetrate the ecosystem that easily,” Prakash explained, “and as it turned out, it could.”
There is something disarmingly ordinary about how the experiment was constructed. Prakash said that he made the website with artificial intelligence, that it was a solo operation, and that it never received money from anyone. He wasn’t a shadowy foreign operative or a rich political consultant setting a trap. He was a 21-year-old person with a laptop, an idea, and enough understanding of the media to know that appearance can matter more than verification. The statement on Median Strategies slick. It insisted that “no individual involved with Median Strategies received compensation or financial benefit from the project,” and repeated that the operation had not been funded, commissioned, or directed by any campaign, political party, media outlet, or outside entity. When asked whether he had quietly taken advantage of the false data by betting on prediction markets, Prakash said that he had not: “No, not at all.” The entire episode seems to have been less a cynical grift than a strange and disorienting curiosity project, the kind a technically talented young person might run to test how brittle public trust is when it is packaged in the right style. He did not have a big machine: as of Thursday morning, the Median account on X had about twenty followers. The account, according to reporting, had only begun posting on August 9. Yet in that short time, it achieved results that many genuinely organized media companies not built.
The realism of the fake polls was perhaps less important than the willingness of institutions to receive them. One surprisingly high poll released by Median appeared to show the mayor of Los Angeles, Karen Bloom Bass, far ahead of her challenger by more than ten points, and why the mayor’s office seized on the data and presented it publicly. This is what makes the digital information ecosystem so frightening. The same poll was picked up by the California Post. A widely followed polling aggregator on X also shared a Median poll from Nevada, despite the fact that several credible polling organizations had already refused to include Median data in their averages, saying that its methodology could not be verified. These are not dark-Russian bots. They are institutions and movement structures: a politician, a news organization, a known aggregator, each hoping to offer something valuable. The news filled a gap for them: it validated the decision they already wanted to make. Bass’s team wanted to show momentum, the California Post wanted a headline, the aggregator wanted to widen its database, and all of them missed what was underneath because the surface was friendly. There were also other fictional polls: that she said suggested that Nevada’s governor, Democrat Joe Lombardo, had risen by five points; another, in Wisconsin’s Democratic primary, that appeared to show a candidate named Francesca Hong in the lead by over 20 points. Polling like that would not just produce misleading headlines — if accepted widely enough, it could actually influence campaigns, donors, voters, and decision-making tables.
Prakash told the Guardian that his project was inspired by frustration with the gap between polls and actual election results. He recalled a Michigan Senate primary in which pollsters had given a major candidate a huge lead leading into election day, but the actual vote came far closer. The gap unsettled him. “This is what inspired me to test my theory,” he said. In the process, he had perhaps not expected to see how quickly something made up could travel through the same unseen infrastructure that carriers real polling data every day. “I was a bit shocked, perhaps at how fast this kind of information can spread,” he admitted. He also expressed regret about the consequences of his effort, apologized directly to the people and campaigns that had seized on his fabricated data. “I’ll apologize straight out to the campaigns of Karen Bass, Nithya Raman and Francesca Hong for this,” he said. It is not the sort of statement one usually hears from a digital manipulator who has been caught; there was no attempt to evade, no hidden smirk, and no systematic distortion. This made it all the more uncomfortable. Media and political systems are normally not victimized by evil masterminds, but by people who are willing to make only a small cut and watch how far it goes.
The Guardian’s investigation into his identity is only as messy as a typical internet search, but it reveals how fragile anonymous traps often are. Prakash did not leave technical crumbs in any obvious way, yet digital traces from the domain registration connected Median Strategies to a network of names and servers. By reverse-searching two domain nameservers tied to the website, the newspaper obtained a list of thousands of other domains using those same nameservers. From all thousands, the one hundred fifteen domains used the same two nameservers, and just one attacker striking that, was also using the same domain name registration service as Median Strategies. That one domain no longer led to an active site; on screen there was only a 404 GitHub error message. But according to additional web records, the domain was tied to a GitHub profile, and inside that profile-site there was a file with a name: Rahil Prakash. It was the kind of connection that a careful internet user might have missed, but for the kind of story that does not require spies or encryption or cyber-espionage. It is the same story as everything else in the modern web: a false face has breathable walls, and authority can be created by design choices rather than fact. What is one person with artificial intelligence can build in an evening of curiosity, an organization with resources and strategy could build in a thousand times more.
The American Association of Political Consultants was one of the first voices to call out the ethical failure of the experiment. It rejected the “social experiment” framing with a stern word. “Publishing fabricated survey data as independent research is not a social experiment,” the group said. “It is a deception directed at voters, at the press, and at the campaigns that rely on public data to make decisions.” That rebuke matters not only because fake polling can fool people one day; it matters because it damages the public’s ability to know what is real the next day. If a pollster appears and releases a survey, no matter how behind the scenes it is plausible, refused by outlets and omitted by aggregators, and soon citizens are taught to doubt real data as well. When Prakash activity finally ended, the wide last route slipped into memory. But he used its defense and ended with the same powerful lesson he had seen through his own slow wondering. “I think my overall takeaway was,” he said, “if you just make it look a bit pretty, it’s rather easy for it to spread.” His work added a more serious concern: if one person could do this without money, without machines, without a team, what could a real organization, a well-funded campaign, perhaps in another country, do in the future? That uncomfortable question no longer belongs only to pollsters. It is addressed to all people who consume politics and information in the age of the screen.
Perhaps the most human piece of this story is that Prakash seems to be have all the qualities that made him a perfect messenger: no wishful, little self-doubt, but also no organized bad intentions. He did not wish to make money, did not use the data for betting, did not have an army, and did not set out to ruin anyone’s day. He was just a person who noticed a flaw in the machine and decided to press it, never thinking the answer would come back so quickly. That should comfort us somewhat, because his wake follows almost nothing to protect vitamin future. But in the strange way that in such matters, this also makes his assertion more dangerous. The next person could be not a 21-year-old experimenter but a powerful government agency or a politician with millions to spend, and the same mechanisms that once allowed Median Strategies to affect Karen Bass’s months could allow a stationed deception to affect a real election. The story is not just about one lie. It is a warning about accelerators: AI, partisan news, niche social media communities, aggregation tools, answer-hungry candidates. The final answer is not that fake polls work. The story is that he is falling into the wrong hands, and that the only barrier between a fake poll and a chairman’s tweet is the willingness of newsrooms and campaigns to check a methodology, ask a direct question, and wait for someone to prove they are real. Prakash shut down Median, apologized to the people he misled, and goes on with his life. But the truth he revealed still has no easy solution: a single person can create the appearance of widespread certainty, and in a tired and watching a moment, a person just might be enough.
Rahil Prakash is 21 years old, a recent college graduate, and the unlikely person behind a strange political trick that worked far better than he thought it would. The Guardian identified him as the creator of Median Strategies, an anonymous website that produced fake election polling in Wisconsin, Nevada and California, and that briefly inserted itself into a surprisingly serious part of the political conversation. The website looked like a legitimate polling operation: it had the right language, the right formatting, the right certainty. It told people what they wanted to hear. Karen Bass, the mayor of Los Angeles, touted one of its polls after the result seemed to show her comfortably ahead of her challenger. A California Post cited the same poll. A widely followed polling aggregator on X picked up a Nevada poll and shared it with the site’s audience. In reality, the whole operation was the work of one person with a laptop, an artificial intelligence tool, and a theory about how broken the public’s appetite for information had become. Prakash did not sound like a malicious political operator when he explained himself to the Guardian. He sounded like somebody who had run an experiment in a laboratory and was honestly astonished to see the results. “I wanted to see if fake polls could really penetrate this ecosystem easily,” he said. “And as it came out, it could.” did not expect all the attention. “I did not expect,” he added, “that the California political press would pick it up on your own.”
His only public statement, posted on the Median Strategies website before it withdrew from the internet, was defensive but quiet. It described the project cautiously as a “short-term social experiment,” and said its purpose was to test whether fake polling “could enter and spread through the political information ecosystem without independent verification.” That phrase had a certain analytical coldness, but the reality is more human and more confusing. After the questions started coming from a Los Angeles Times reporter, the site pulled down its material and disappeared almost overnight. Prakash said he built everything with artificial intelligence and was the only person involved. He was not trying to persuade the voters, at least according to him, and not trying to get rich. He insisted that no one in Median Strategies had received any compensation, that no campaign, party or particular media interest had instructed him, and that he had used the fake data as a kind of test rather than an actual attack against anyone. “Did I expect it to accelerate in the way it did?” he asked. “Most definitely not.” The word “fake” may suggest a crime or understood malice, but Prakash seemed to see his work more like a prank with a sample size. He wanted to know whether the shield had holes. These holes were big enough that an entire project could pass through.
What maybe more disturbing than the lies themselves is how quickly they were swallowed. The median poll for Mayor Bass’s race in Los Angeles appeared to show the mayor with a wide lead, and that made it a warm piece of news for a campaign that wanted supporters to see promising numbers. The poll looked polished, and no one could at first notice any obvious contradiction. It was essentially the same in Nevada, where the fictional result was said to show Governor Joe Lombardo at the top; likely the exact kind of story that could change the way a general election is discussed. At least one known polling aggregator on X took the bait and shared the Nevada result with its followers, though only after some credible polling specialties had quietly ignored the numbers because they could not find the methodology of Medium Strategies. This is a remarkable lesson: while professional careful organizations did not accept the data, the most politicized parts of the system did. A politician, a small publication, and an aggregator are exactly the three kinds of actors who are commonly expected to protect the public from false information. Instead they spread it. This had real consequences in no measurable ones: a candidate could have changed their strategy by trusting the wrong number. More importantly, the structure of the election public trust was exposed. If a person with no special source file and twenty followers on social media could create a illusion, then something fundamental in the public conversation has clearly shifted.
Prakash explained that he began something in part because he had been interested by a reality in Michigan, where polling had shown a candidate with a heavy advantage before the primary, but the actual result was much closer. That difference appears to have unsettled him. Rather than write a paper or discuss the problem in a classroom, he decided to prove to himself that the system was defective. And so he tested it. He apparently used a normal domain registration setup, built a device that looked durable, and made only simple lines to that the presentation glamor. “I think my overall takeaway was that if you just make it look somewhat beautiful, it’s rather easy for it to spread,” he said. The statement suggests something deep about how modern political information is produced: often what matters more than factual accuracy is the veneer of credibility. If a poll comes in a attractive chart, with a name like median strategies and a message that supports the existing strategic agenda, someone will share it. From Practically’s point of view, this was less a matter of fraud than a proof of concept. He never anticipated the sudden and public attention. He also spelled a true honest apology: “I’ll apologize directly out to the campaigns of Karen Bass, Parsmy Raman and Francesca Hong for this.” That is an important note in a story that could have been part of a normal political manipulation. He recognized that these real people and their campaigns had been affected by the product of his curiosity.
The Guardian’s discovery of Prakash was not the kind of detective work one might expect from a thriller; it was digital forensics built one link after another. The newsroom looked at the domain registration records for Median’s hi. By searching through two nameservers connected to the site, it found a group of thousands of domains. Among those, roughly one hundred and fifteen had used the same two nameservers, and a single one also used the same registrar as Median Strategy. That domain no longer led to a dynamic; instead, it displayed only a GitHub 404 error. But a set of connected records connected that domain to a GitHub profile, and in that profile someone had left a file with the name “Rahil Prakash.” That was enough to make the connection. This was not an exotic piece of digital espionage. It was the work that any determined reporter could complete, because the digital space still leaves footprints. The quiet surprising part is that someone with a technical enough perspective could hide so openly: the media organization was not real, but the data of its construction was real enough to be followed to an actual human. And when they finally contacted that human, he did not hire, did not ask for a global PR conference, did not create more documentaries. He simply answered and confirmed what he had done.
At one point, the effort was finally disbanded completely by those who understand the public value of real survey. The American Foundations Association named it clearly: “Publishing fabricated survey data as independent research is not a social experiment.” Its phrase “a deception directed at voters, at the press, and at the campaigns that rely on public data to make decisions” combines the entire problem in one sentence. The risk is not just the insult to the campaign; one may make safe one lie, but it contributes to a deeper erosion, a general distrust, a future in which the voters can no longer know what to trust. The median themselves finally said: “If one person can do this, what can a real operation do in the future — a real operation by an organization, a country, whatever?” That is the question that now remains in the war. Prakash, twenty-one years old, has moved from his experiment. He will not make money; his site is closed; the attention will eventually be refused. But the question he began cannot be erased. Perhaps the lesson is not that political information is worse than we thought, but rather that,the production of truth depends on people who are not deceived by appearance. The world no longer has any shortage of people who build fake information; the world needs enough people who insist on asking. If a fake poll could spread this easily, then the real defence of the different ecosystem is still the old news. Human verification, methodical check, and the willingness to wait before sharing something that looks player but is not real.