A second polling firm is under fire after admitting to fabricating data in a key 2026 primary election, this time in Florida.

The polling monitor FiftyPlusOne wrote in a blog post on Thursday that it had received a tip suggesting that pollster The Public Sentiment Institute (TPSI) had violated one of its rules by not disclosing publicly that a survey it had released of the Republican gubernatorial primary field was commissioned by one of the candidates — a disclosure pollsters are required to make by most polling aggregators to be considered legitimate firms, as campaign-sponsored polls often tilt in the favor of the relevant candidates.

But in the course of investigating that tip, FiftyPlusOne says its analysts discovered a second problem — that a TPSI poll postmortem published in the wake of the Florida primary election on Tuesday admitted TPSI pollsters had changed the results of some respondents’ answers to lean towards James Fishback, the candidate who’d sponsored the poll. Fishback went on to be defeated in Tuesday’s primary, placing third behind Jay Collins and the winner, Rep. Byron Donalds.

“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,” read TPSI’s postmortem.

FiftyPlusOne called the admission evidence of an “obvious violation of standard polling practices,” which it said was “made all the more questionable by the undisclosed relationship between the pollster and the Fishback campaign.” The polling aggregator said it would not use TPSI polls in its analytics in the future.

The admission from TPSI is one that will further challenge trust in polling firms and election analytics at a moment when that trust is already badly shaken. The news comes just days after a group calling itself “Median Strategies” announced that it was shutting down after publishing fake poll results in the Los Angeles mayoral race as part of a “social experiment”.

“All previously published polling releases have been withdrawn and should not be cited or treated as genuine polling data. Median Strategies was created as a short-term social experiment examining how purported polling information could enter and spread through the political information ecosystem without independent verification,” read the group’s statement. A 21-year-old recent college graduate was revealed by The Guardian as the owner of the site and fake company, whose “polls” were shared widely online after they showed two candidates in two separate races leading their opponents by wide margins.

“I wanted to see if fake polls could really penetrate the ecosystem that easily,” Rahil Prakash said in an interview with The Guardian, adding: “And as it turned out, it could.”

Prakash’s fake polling data gave fans of Francesca Hong undue confidence ahead of her defeat in last week’s Wisconsin gubernatorial primary, and was even cited by the campaign of Los Angeles Mayor Karen Bass as she runs for re-election against progressive Nithya Raman. They were not picked up by credible polling aggregators, but the two polls spread quickly online anyway.

And the 2026 cycle follows an embarrassing cycle for legitimate pollsters; the 2024 presidential election was capped off by a poll by much-cited Iowa polling guru Ann Selzer which showed Democratic Vice President Kamala Harris in the lead in Iowa with only days to go before the election. Harris went on to lose not just Iowa but every other major battleground state in her matchup with Donald Trump, and Selzer retired from political polling. She insisted afterwards that her decision was not related to the serious discrepancy between her final poll and the election results.

Polling this cycle looks to be little better. In Michigan, the industry’s collective margin of error in the highly-watched Democratic primary for U.S. Senate was more than 10 points, according to an analysis from local newscaster WLNS 6.

Though Prakash didn’t draw a conclusion from the results of his “social experiment”, it’s safe to say that the trust issues surrounding political polling in the U.S. will only be made worse by further admissions like the ones this week from TPSI.