Political polling in the United States is shifting from a tool used to measure public opinion into a weapon used to manipulate it, according to a detailed report by The New York Times. A wave of unreliable data, fake surveys, and volatile prediction markets is warping campaign narratives, misleading donors, and distorting news coverage like never before.

While Donald Trump and some Republicans have historically used dubious polling to advance their agendas, the phenomenon has recently exploded within Democratic primaries, creating an air of false reality that blindsides both candidates and voters. The industry's accuracy problem is not new — polls understated Trump's support in both 2016 and 2020, drawing formal post-mortems from the American Association for Public Opinion Research, the field's main professional body — but the sheer volume of cheap, unvetted surveys now in circulation, and the betting markets attached to them, have multiplied the paths bad numbers can travel.

Michigan and Wisconsin offered real-world proof

The consequences of this polling distortion were on full display during recent midterm contests in Michigan and Wisconsin.

In Michigan's Democratic Senate primary, consecutive public polls showed progressive candidate Dr. Abdul El-Sayed leading moderate Representative Haley Stevens by double digits. The narrative of an impending blowout was so pervasive that reporters openly questioned Stevens about her opponent's massive lead on primary day, and El-Sayed's campaign was confident enough to celebrate early poolside. El-Sayed ultimately squeaked out a 1-point victory, proving the double-digit blowout predicted by the polling data completely wrong.

A similar outcome unfolded in Wisconsin's gubernatorial primary. Polls consistently painted a picture of inevitable victory for Francesca Hong, leading heavyweights such as Mandela Barnes to drop out early. When David Crowley won the nomination instead, the result was treated as a shocking upset — largely because the public had been conditioned by flawed data to believe Hong was unbeatable.

A structural collapse in the polling industry

The breakdown in accuracy is fueled by a structural collapse within the polling industry itself. Traditional, high-quality "live caller" phone surveys have become too expensive for most firms to sustain. In their place, the industry has turned to cheap online surveys and text-message polling — an issue PBS NewsHour examined in its reporting on why primary polls appear so inaccurate this year. While affordable, these digital methods heavily over-represent younger, tech-savvy demographics while missing the older voters who actually show up at the ballot box. That methodological vulnerability has turned the political landscape into a "Wild West" largely devoid of regulation.

The problem has grown severe enough that completely fabricated polls have slipped into the mainstream. The Los Angeles Times recently exposed three entirely fake surveys — covering the Los Angeles mayoral race, the Wisconsin gubernatorial primary, and the Nevada governor's race — that were injected into the media ecosystem as a self-described "social experiment." For readers, the field's own standards offer one rough filter: pollsters in AAPOR's Transparency Initiative commit to disclosing their methodology, sponsors, and sample sizes — exactly the kind of details that fabricated or low-quality polls cannot credibly provide.

Prediction markets amplify the chaos

Compounding the problem is the explosive rise of political prediction markets. Platforms such as Kalshi and Polymarket allow users to bet on election outcomes, creating powerful financial incentives to manipulate data. Kalshi operates as a Commodity Futures Trading Commission-regulated exchange whose election contracts were cleared by a federal appeals court in 2024, and Polymarket reached a 2022 CFTC settlement over offering event contracts without registration — but that oversight governs trading conduct, not the accuracy of the outside polls that feed the odds. Because these betting odds are highly sensitive, campaigns can flood the internet with cheap, favorable polls to artificially spike their candidate's odds. Major media organizations then amplify these skewed metrics through lucrative partnerships, displaying gambling probabilities on live television and in print as if they were legitimate reflections of voter sentiment. Related reporting has examined Trump and prediction markets as well as insider trading concerns on prediction markets. The result is a feedback loop in which artificial momentum dictates donor behavior and strategic voting. One open question is how newsrooms will label such odds going forward, and whether they will keep disclosing the pollsters and sponsors behind the numbers they average.

As the 2028 presidential primary season approaches, political strategists warn that every candidate will likely deploy their own "cooked poll factory." That prospect leaves major political donors and ordinary citizens in a state of profound confusion, unable to separate genuine popular support from manufactured digital illusions. When the empirical data used to guide a democracy becomes untrustworthy, the purest element of the system remains the physical act of voting itself.