A $20 bill that a bank teller miscounts is still a $20 bill. A $20 bill printed in a basement is something else. Median Strategies appears to have done the second thing to two elections in one week, inventing the voters, the interviews, the methodology, and the results, then posting the product under a corporate-sounding name the day before Wisconsin Democrats chose a nominee for governor. Calling the stunt a "social experiment" after it was exposed does not convert attempted election fraud into political science.
On August 10, one day before Wisconsin's Democratic gubernatorial primary, an anonymous shop calling itself Median Strategies published a supposed survey putting Francesca Hong at 46.3% and David Crowley at 22.6%. The release came with a sample, field dates, a methodology statement, and a name that sounded like one you might already have heard of. Someone reading it on a phone the night before the primary would have seen the ordinary furniture of a real poll and had no reason, from the face of the document, to treat it as fiction. The poll did not exist.
Crowley won the August 11 primary 39.8% to Hong's 39.3%, a margin of 3,796 votes. Median had overstated Hong's advantage by more than 24 points. That gap is not a polling miss, because no voters were interviewed and no sample was ever in the field.
Median did not confess when the returns came in. On August 12, it behaved like a legitimate firm that had simply gotten Wisconsin wrong "by quite a margin," the kind of rueful note a real shop posts after a blowup, the kind of note that keeps the brand intact for whatever it planned to publish next. The next release arrived the following day.
On August 13, Median published a second fictional survey, this one from Los Angeles, claiming Mayor Karen Bass led Nithya Raman 50.18% to 38.39% among 560 supposed likely voters interviewed July 30 through August 5. The Bass campaign promoted the number and then deleted the post. A newspaper reported it. Spencer Pratt amplified it to thousands of reposts and tens of thousands of likes. A mayoral race absorbed a figure that had never been collected, and the people who moved that figure along treated the corporate name as a substitute for a phone log.
After repeated questions from the Los Angeles Times, Noah Goldberg also carried the questions onto 𝕏, and Median admitted on August 17 that the polling was fake, a "short-term social experiment" examining how fake polling "could enter and spread through the political information ecosystem without independent verification." The operators vanished from the site, and an unsigned statement withdrew all previously published releases and stated that none of it should be cited as genuine. The Guardian identified the person behind the shop as Rahil Prakash, a 21-year-old recent college graduate who owns the site and the fake company. Reached by phone on Thursday, Prakash told the Guardian, "I wanted to see if fake polls could really penetrate the ecosystem that easily." He added, "And as it turned out, it could." The Guardian said the fakes ran in Wisconsin, Nevada, and California. The website named no pollster, no principal, no funder, and no legal entity. The domain had been registered on August 10, the same day the Wisconsin release appeared.
No public evidence implicates Hong, Crowley, Bass, Raman, or their campaigns. Prakash deceived them too. That fact does not make the campaigns look careful. Democratic operations that boost a number from a shop with no named pollster, then delete the post after the number collapses, are telling you how they treat evidence when the evidence is flattering. The invented numbers spread online after they showed two candidates leading by wide margins, and they gave Hong's supporters a confidence the ballots did not justify.
Wisconsin already makes it illegal to knowingly publish a false claim about a candidate to influence votes. Section 12.05 of the Wisconsin Statutes forbids any person from knowingly making or publishing, or causing to be made or published, a false representation pertaining to a candidate or referendum which is intended or tends to affect voting at an election. Median published a false survey about Hong and Crowley on the eve of the primary, gave Hong a 23.7-point lead built from interviews that never happened, and later admitted the fabrication was deliberate.
The statute is written in the alternative. Wisconsin does not ask whether a single voter flipped. The law is satisfied if the lie was meant to move votes or if it tended to. Median's own explanation supplies the intent. Prakash said he wanted to see whether fake polls could penetrate the ecosystem that easily, which meant reliance was the metric he chose.
The Supreme Court has held that falsity alone does not strip speech of protection. This case does not depend on punishing every political lie. Wisconsin's law applies only to a knowing false representation about a candidate, tied to voting. "Hong is going to win" is a forecast. "We interviewed 560 people and 46.3% chose Hong" is an assertion about events that did not happen. The deception concerned the existence of the evidence, not its interpretation. Fraud doctrine does not treat forged instruments as ordinary opinion.
California has its own statutes, and they are written for people who try to cheat a vote, not for people who call a candidate a name. Elections Code 18500 makes it a felony to commit or attempt fraud in connection with any vote cast, to be cast, or attempted. Section 18502 punishes anyone who in any manner interferes with voters exercising their rights so as to prevent a fair election. Those provisions have generally been applied to ballots and registration rather than campaign speech. Median did not trade in insults. It constructed a fictitious research institution, invented 560 voters, assigned them preferences, fabricated field dates and a margin of error, and released the product into an active mayoral election.
A material representation went out under a firm name. Prakash knew it was false. He intended for others to rely on it, and a campaign, a newsroom, and large accounts on 𝕏 did so. A confession after exposure helps prove the attempt.
A legitimate study informs its participants, obtains consent, and discloses its methods. The Heritage Foundation and other election-integrity researchers insist that the process that produces political information is part of the public trust. Median secretly conscripted voters, campaigns, journalists, and traders as unwitting subjects in elections that counted. Nobody who voted in Wisconsin or Los Angeles was told they were part of a study, and nobody at a campaign or in a newsroom was asked for consent.
I wrote a few weeks ago about Texas Public Opinion Research, a single survey showing James Talarico leading Ken Paxton by 5 points, a Texas Tribune headline, and a Talarico campaign that promoted the result as if a nonpartisan shop had discovered it. That poll was not fictional. The interviews happened. The misrepresentation was institutional. TPOR called itself nonpartisan, even though its director is a founding partner of a fund created to help Democrats win in 2026, which had paid nearly $200,000 to the consultancy conducting TPOR's polling. TPOR laundered partisanship through that word. Median laundered fiction through the word "poll." The Talarico operation used real respondents and a borrowed costume.
The same week Median was being unmasked, a second polling firm admitted fabricating data in the Florida Republican gubernatorial primary. FiftyPlusOne got a tip that The Public Sentiment Institute did not disclose that a survey of the GOP gubernatorial field was commissioned by one of the candidates, a required disclosure because campaign-sponsored polls often tilt. While investigating, FiftyPlusOne found TPSI’s postmortem after Tuesday’s Florida primary admitted the firm changed some respondents’ answers to lean toward James Fishback, the candidate who sponsored the poll. Fishback lost, placing third behind Jay Collins and the winner, Rep. Byron Donalds. TPSI wrote that “Jay Collins was polling in the twenties in our data. That did not align with our read of the race, and we concluded it might reflect a data-quality problem specific to our text-message panel. Acting on that conclusion, we identified respondents who selected Collins but whose broader profile aligned with a competing coalition, and leaned those respondents toward James Fishback.” FiftyPlusOne called that an “obvious violation of standard polling practices,” “made all the more questionable by the undisclosed relatoinship between the pollster and the Fishback campaign,” and said it will not use TPSI polls in its analytics in the future. Median invented the voters it published. TPSI moved the ones it already had.
Median did not cost Hong the primary. Legitimate pollsters misread Wisconsin the same way. State Navigate had Hong at 22%, down from 44%. Marquette had Hong at 38% against Crowley at 7%. Markets had Hong near 95% before the fake poll could plausibly have mattered. A reader watching Navigate and Marquette would already have believed Hong was running away with it, which is why the fake could travel. Median copied that consensus. The invented result appeared credible because respected firms had publicly reported similar figures.
Roughly $19.83 million in cumulative volume sat on the Wisconsin primary markets on Polymarket; traders were digesting polls, endorsements, coverage, and each other. The cascade ran poll to headline to campaign post to social proof to more coverage and repetition replaced verification. There is no evidence Median shifted those odds. The question still belongs to investigators rather than to assumption.
The remedy is not new legislation. Wisconsin prosecutors should investigate 12.05. California authorities should examine 18500 and 18502. Preserve the poll documents. Subpoena the domain registration and the hosting. The Guardian has identified Rahil Prakash. Investigators should still confirm that he controls the 𝕏 account and the domain, and determine whether he or anyone else traded on information he knew to be fabricated. Prakash claims no campaign, party, PAC, or financial interest funded the work. That claim merits the confidence the polls merited, which is none.
Median proved that polling is political power, that labels substitute for evidence, and that a manufactured number travels before anyone asks if the patient existed. The plain name for what they did is an attempt to manipulate elections with counterfeit evidence. Prosecutors have the statutes; we now have a name rather than only an 𝕏 account with an unknown owner, and the counterfeit is still sitting in the archive.
If you enjoy my work, please subscribe https://x.com/amuse/creator-subscriptions/subscribe
Alexander Muse is a Fellow at the John Milton Freedom Foundation and publishes daily political analysis at amuseonx.com. Primary sources cited in this piece are linked inline; campaign finance figures are drawn from FEC filings, polling data from publicly released crosstabs, and legal claims from filed pleadings. Corrections are posted to the original URL with a dated changelog. Readers who identify errors are invited to contact the author directly. Each op-ed edited for grammar and clarity using Ai in a sponsored partnership with Grammarly. Data provided in a sponsored partnership with Polymarket.
I have never used polls as a guide in casting my vote. I can, however, see that polls may be more consequential in states like California that use a top two open primary system. I can also see that Fake polls can influence online political betting markets, so that their purveyors can make a quick buck at the expense of those they’ve tricked. In states with large populations of ESL legal voters, I believe polls may also have greater impact. This is an important issue as we pursue cleaning up our voter rolls, establishing voter identification, and putting integrity back into our election system.
Counterfeit polls belong in the same category as counterfeit ballots, forged petitions, and fabricated evidence. The objective is identical: distort reality before voters make a decision. Democracy survives disagreement. It cannot survive a marketplace where fabricated evidence is treated as legitimate campaign speech until someone gets caught. Muse identifies the real danger. The scam worked because campaigns, reporters, influencers, and traders rewarded the appearance of credibility instead of verifying it. That is the upstream problem. Every fake poll poisons the information ecosystem before a single vote is cast. Prosecutors should treat deliberate counterfeit polling as election fraud, not political theater.