Fake Polls, Phantom Experts, and the New Age of Fabricated Reality

The media world has been thrown into fresh consternation this week by the discovery that fake polls from a nonexistent pollster had been cited by major news organizations and polling aggregators, including The New York Times itself. Government officials immediately promised swift and legally dubious punishments for broadcasters that use such fabricated data, but the scandal is not simply a cautionary tale about one bad actor. It is part of an ongoing pattern of hoaxes, “social experiments,” invented personas, and AI-generated content that has turned large swaths of the information ecosystem into something resembling a hall of mirrors. The latest episode centered on Wedgewood Polls, an outfit exposed when FivePlusOne, an outlet investigating the pollster, contacted Wedgewood for a comment. In response, Wedgewood posted a statement, later deleted, declaring that it was an “experimental project to demonstrate the vulnerabilities, lack of oversight, and overall barrier to entry in the modern polling industry.” The fake pollster went on to claim that it was “far too easy” to be taken seriously in media and public commentary. The explanation was immediately met with suspicion, not least from the prominent pollster Nate Silver, who was very skeptical that the operation was really a social experiment rather than a convenient cover story for a deception that had gotten out of hand. But whether Wedgewood was run by academics, provocateurs, or engagement seekers, the damage to public confidence in polling had already been done. The episode also echoed a similar incident in August, when the Los Angeles Times reported that Median Strategies, another purported polling firm, had produced a bogus poll showing Los Angeles Mayor Karen Bass leading Councilmember Nithya Raman by a wide margin. Earlier, the same firm had released a phony poll showing Francesca Hong, who ultimately lost her race for Wisconsin governor, up by more than 20 points.

Median Strategies told the Los Angeles Times that the firm was “created as a short term social experiment to examine how purported polling information could enter and spread through the political information ecosystem without independent verification.” If these operations were experiments in the credulousness of modern journalism, they succeeded beyond what their creators likely imagined. Nobody is completely immune to deceit. Savvy media consumers have become increasingly attuned to detecting AI-generated writing, deepfake videos, and algorithmically manufactured images, yet masses of people still fall for fake polls, fake experts, and fabricated statistics. The Wedgewood affair is only the latest reminder that polling, despite its mathematical pretensions, rests on trust. A poll’s credibility depends on the assumption that the organization conducting it is real, that its methodology is sound, and that its results are honestly reported. When those assumptions fail, the damage is not confined to one bad election cycle. It corrodes the very idea that public opinion can be measured, and it gives politicians an excuse to attack the media for publishing numbers they do not like. The government response has been predictably heavy-handed, with talk of legal punishments for broadcasters. But the deeper problem cannot be solved simply by fining news outlets. Wedgewood was able to infiltrate the ecosystem because too few editors, reporters, and aggregators asked basic questions before treating its output as newsworthy. The same failure has manifested across the industry, from confident citations of unverified polls to articles built around sources who do not exist.

This is not a new problem. In 2018, when the word “slop” still meant food for pigs and long before anyone was writing posts with ChatGPT, Intelligencer asked a direct question: “How much of the Internet is fake?” The answer, according to the outlet, was “a lot.” Reason editor Nick Gillespie, reflecting on the article at the time, described it as a “sobering and obvious revelation that much of what we browse, engage, and interact with is in some fundamental way, total bullshit.” Gillespie observed that the internet was not the disintermediated, utopian world celebrated in the early days of Wired under Louis Rossetto and Jane Metcalfe. Instead, it was something closer to the endlessly creative, weird, off-kilter, semi-insane world imagined by Philip K. Dick and, later, Neal Stephenson. That observation has aged remarkably well. Eight years later, the semi-insane world is no less insane, and artificial intelligence has lowered the cost of producing fabricated content to almost zero. The term “slop” has been repurposed to describe the endless river of low-quality AI-generated articles, images, videos, and social media posts that now flood our screens. In this environment, human beings, as they always have, continue to fall for increasingly intricate schemes and hoaxes. The problem is not merely technical. It is social. People are more likely to trust information that seems familiar, that comes from an authoritative-looking source, or that confirms their preexisting beliefs. That has always been true, but the scale and speed of the modern internet make the consequences far more dangerous. A false poll can be circulated to millions within hours. A fake expert can appear on multiple news programs before anyone checks the facts. A single YouTube video can shape public perception of an election, a health crisis, or a war.

Recent reporting shows how pervasive the deception has become. In September, Semafor’s Max Tani revealed that a large number of YouTube videos—several of which had garnered hundreds of thousands of views—were made by “ghost creators.” These videos were presented as news reports, but real actors had simply been paid to read AI-written scripts that were laced with misleading claims and outright falsehoods. According to Tani, the accounts behind these videos appeared to be run not by political operatives trying to win an election, nor by foreign intelligence services seeking to destabilize the United States, but by “engagement junkies detached from the stakes of politics” who were simply chasing views. While any semi-media-literate person in 2026 knows that not everything on YouTube is true, the scale of the operation was staggering. These videos looked and sounded like legitimate journalism, and they reached enormous audiences. The same story has unfolded in the world of expert sources. Qwoted, a platform that connects journalists with sources, was forced to remove a supposed art therapy expert from its website after her details were found to be false, according to the Press Gazette. By the time the deception was uncovered, the fake therapist’s insights had been cited in several outlets, including Psychology, Forbes, Glamour, and Vice. Even polling data can be laundered through dubious companies. The original commentary raised concerns about news outlets and commentators citing polls published by EduBirdie, a Cyprus-based company that sells human-written academic papers to students for $8 a page. It is, ethics aside, a bold business model in the AI age, yet the fact that its polls were treated as credible only reinforces the sense that many media organizations are still not doing the basic work of source verification.

So what is to be done? As Nick Gillespie wrote in 2018, there are existing legal levers to root out fraud, but any top-down approach to squashing online lies “will be poorly applied, repressive, and ultimately arbitrary.” The experience of government-led crackdowns on COVID-19 misinformation proved him right. When governments begin to decide what is true and what is false, they rarely stop with actual misinformation. They suppress dissent, punish uncomfortable reporting, and create a chilling effect that makes everyone less informed. Outsourcing the arbitration of truth to a single authority would only diminish access to information, and no one entity—particularly the government—should be trusted to determine what is true or false. Unfortunately, the Trump administration has not learned much from the pandemic era. It has instead repeatedly tried to crack down on media coverage it does not like. The instinct to use state power against falsehood is understandable, because the scale of misinformation can seem overwhelming. But the history of such efforts is uniformly discouraging. Legal pressure tends to fall on smaller outlets and independent journalists, not on the well-resourced operations that produce the most damaging lies. Meanwhile, the government itself has often been a source of misinformation, making it an especially poor arbiter of truth. The solution cannot be found in censorship, fines, or political pressure. The solution must be found in the institutional culture of journalism itself—but that culture has shown little sign of changing. Media organizations are supposed to act as B.S. detectors, checking facts, verifying sources, and refusing to repeat unsubstantiated claims. All too often, they fail.

The burden therefore falls on the individual consumer. It is a difficult and sometimes exhausting task, but it is unavoidable in a world where fake polls, ghost creators, and invented experts have become routine. Consumers must learn to wade through lies, AI hoaxes, and self-described social experiments without losing their sanity. They must check the source of a poll before sharing it. They must ask whether a quoted “expert” can be verified through multiple independent channels. They must be suspicious of articles that seem perfectly designed to confirm their own views. This sleuthing for scams may even be enjoyable. As the original commentary notes, why pay to have your brain scrambled by Nathan Fielder’s Elizabeth Holmes documentary when you can drive yourself insane by tracking down and scrutinizing widely cited statistics and polls? It is a wry observation, but it points to a serious truth: the same digital infrastructure that makes large-scale deception possible also gives ordinary people the tools to detect it. A search engine can trace a suspicious statistic to its original source. A quick check can reveal whether a polling firm has a track record, a methodology, or even a real address. The more people practice this kind of skepticism, the harder it becomes for fake pollsters and ghost creators to succeed. There will always be new scams and new forms of manipulation. But the core requirement remains unchanged: verify before believing. Without that discipline, we are left in the semi-insane world described by Gillespie—a world where authority is whatever manages to get published, and truth is whatever manages to get repeated. That would be a terrible outcome for journalism and for democracy alike. It is also, increasingly, the world that we have already made for ourselves.