A Fake Polling Company Got Its Numbers Into Major News | RMN
Not viral. Useful.

A Fake Polling Company Got Its Numbers Into Major News Outlets. Then It Called the Whole Thing an Experiment.

Wedgewood Polls invented election surveys that entered national coverage and influential polling trackers. Its admission exposed more than a weakness in election forecasting: it showed how easily credible-looking information can borrow the authority of institutions that repeat it.

· · United States

An election poll arrives with the familiar packaging of authority: candidate names, percentages, field dates, a sample size and a margin of error. It looks like somebody asked real people real questions and carefully counted the answers. A reporter can use it to explain a close race. A campaign can point to it as evidence of momentum. A polling aggregator can fold it into a mathematical average. Each repetition makes the original number look more established, although none of those steps necessarily proves that anyone was ever surveyed.

That is the problem exposed by Wedgewood Polls, a purported political survey firm that spent the 2026 midterm cycle publishing results in competitive races before acknowledging on October 20 that nearly all its polls were invented. Its figures had already reached national news coverage and some of the country's best-known election trackers. After investigators challenged the underlying data, Wedgewood described the operation as an experiment designed to expose weak verification in the polling industry.

A deception can reveal a weakness while still being a deception. The interesting question is not whether Wedgewood proved that journalists and forecasters should scrutinize polling more carefully. It did. The harder question is how fabricated numbers traveled through systems designed to help people understand public opinion, and what happened to the people who were expected to believe them.

How a made-up number becomes a public fact

Wedgewood began releasing election surveys in June. Its results moved through the same channels used by legitimate pollsters: a website, methodological descriptions, social-media announcements and releases that could be picked up by political reporters. The Guardian documented examples in Newsweek, The Hill, Breitbart, The Daily Beast and the Washington Examiner. The New York Times displayed 15 Wedgewood surveys on its outside-poll tracking pages; Decision Desk HQ incorporated 17 into its race-level polling lists and averages.

The numbers had different kinds of authority in different places. Inclusion in a news story gave them a recognizable outlet's name. Inclusion in a tracking database placed them alongside work by established firms. Incorporation into an average made them part of a calculation whose result looked more sophisticated than any single survey. This is the information-laundering problem: an unsupported claim can emerge from a series of otherwise legitimate processes looking more credible than it did at the start.

It would be easy to conclude that the outlets simply accepted any poll delivered in a professional-looking format. The record is more uncomfortable. The Times said Wedgewood answered detailed questions about its personnel, funding and survey methods before its results were listed. FiftyPlusOne, a polling analysis organization staffed by former FiveThirtyEight researchers, said it had also asked about the firm, received names, reviewed its methodology and independently confirmed that one person presented as involved had experience in polling. The organization still was not what it claimed to be. Conventional due diligence did not establish that the data itself was real.

The decisive check came from inside the supposed survey records. After receiving a tip with supporting evidence, FiftyPlusOne examined raw data from a Wedgewood Mississippi Senate poll that the firm claimed to have conducted through the online platform Pollfish. Investigators purchased a separate survey using the same provider and compared its metadata with the material Wedgewood supplied. 

Respondent identifiers and transaction identifiers did not behave as they should have; the relationship between a transaction number and when an answer was submitted was especially suspicious. In other words, investigators did not have to guess whether the political result seemed believable. They tested whether the digital fingerprints of a survey looked like those of a survey that had actually happened.

On October 20, after FiftyPlusOne sought an explanation, Wedgewood publicly admitted fabricating its data and framed the episode as a warning about the industry's vulnerability. It claimed one crowdfunded Nebraska poll was genuine, a claim that does not by itself verify that survey. The admission followed the investigators' questions, rather than arriving as the scheduled conclusion of an independently documented research project.

Calling it an experiment does not erase the consequences

Wedgewood's explanation borrowed the language of public interest. The operation had, in fact, demonstrated that false polling could be amplified by important institutions. But a meaningful test of a system's reliability is not automatically a license to insert fabricated evidence into real political contests. Journalists, campaigns and voters did not consent to take part in an exercise in which invented measurements could shape their understanding of live races.

Michigan state senator Mallory McMorrow, who had been running for the U.S. Senate, described a particularly human cost. She said one Wedgewood survey showed her support collapsing, contributing to a loss of donations, distressed volunteers and pressure from reporters. Other polling also showed a difficult race, so her account cannot establish how much Wedgewood alone changed her campaign's trajectory. It does establish why this was not a harmless demonstration contained inside a laboratory: people were responding to the numbers in the world where elections actually happen.

The corrections were substantial. The Times removed its 15 listed Wedgewood polls. Decision Desk HQ removed 17 and recalculated its forecasts, reporting that the changes did not alter its projected congressional seat totals or probabilities of chamber control. Newsweek said it retracted three stories and corrected others, while additional outlets deleted or amended references. That distinction matters. The available evidence does not show that these fake surveys changed an election result or a major aggregate forecast. Their demonstrated damage was to reporting, individual narratives and confidence in the systems that were supposed to filter bad information.

Corrections are necessary, but they face an asymmetry of their own. A fraudulent number can be shared in a headline, captured in a screenshot, repeated in a discussion or syndicated to another website before the original publisher adds an editor's note. The correction may be scrupulous and visible on the source page while the misleading first impression keeps circulating elsewhere. Publishing an error and repairing the record are not equal-sized events in the attention economy.

What New Jersey's polling expert wants readers to check

The Rutgers connection comes through Ashley Koning, director of the Eagleton Center for Public Interest Polling. In an October 30 Rutgers interview about interpreting the 2026 midterms, she specifically cited Wedgewood's admission as a reason for readers to examine where polling comes from. Her questions are direct: Who runs the firm? Who paid for the survey? Who was interviewed? When and how were they reached? What were the exact questions? A respectable-looking percentage is no substitute for answers.

Koning also highlighted why this is especially important for places such as New Jersey's competitive 7th Congressional District. District-level polling is expensive, its samples are smaller, and relatively few independent organizations conduct it regularly. Those gaps create demand for any fresh numbers that appear to describe a close contest. Even a legitimate advocacy poll needs careful interpretation because sponsors may choose to publicize the results most favorable to their cause. A fabricated poll exploits that same hunger for local information without conducting the underlying work at all.

This does not mean that polling itself is worthless. Koning emphasized that a credible poll is a snapshot, not a prediction, and that survey methods can be evaluated. Honest surveys can still be wrong because of sampling, turnout assumptions or changing voter preferences. A made-up survey is a different category of failure. Treating both as proof that nobody can be trusted would punish the researchers who disclose their methods and the journalists who uncover fraud.

There is a broader lesson for local journalism, too. Institutions accumulate credibility by doing work over time: attending meetings, interviewing residents, correcting mistakes and explaining how they reached their conclusions. An imitator can copy the visible format in an afternoon. A fabricated poll can imitate the structure of research just as a politically operated website can imitate the structure of a neighborhood newspaper. The form is not the proof; the reporting and evidence behind it are.

The hardest change may be learning to regard a polished presentation as the beginning of verification, not the end. Before a number becomes a headline, someone has to establish what was measured and whether the measurement exists. Before it becomes a forecast input, someone has to decide whether the source deserves inclusion. And when that chain breaks, the correction should travel at least as deliberately as the original claim.

Wedgewood wanted the public to see how easily invented figures could pass as credible polling. The episode did reveal that vulnerability. But the most important fact about its surveys remains simple: the people whose opinions were supposedly being measured were mostly imaginary. Every institution that repeated the results gave those imaginary respondents a little more public authority. 

That is what the next round of scrutiny needs to prevent.

Sources

Rutgers Research — Ashley Koning on interpreting election polls (Oct. 30, 2026)

FiftyPlusOne — Original investigation and Wedgewood admission (Oct. 20)

The Guardian — Pollster admits fabricating data (Oct. 21)

Axios — Reporting on Wedgewood, scrutiny and industry response (Oct. 21)

The New York Times — Removal of 15 polls, as archived by The Catch (Oct. 21)

Decision Desk HQ — Removal of 17 polls and forecast update (Oct. 21)

Newsweek Midterms Monitor — Retractions and corrections (Oct. 22)

Reporting attributed to the Rutgers, FiftyPlusOne, and related materials cited in Sources. Cultural framing is RMN's.

Editorial standards · Corrections