Wedgewood Polls admits nearly all of its surveys were fabricated after a polling aggregator found impossible-looking raw data. The fake numbers were nevertheless tracked by major media outlets, aggregators and real campaigns.
A polling firm that spent months publishing numbers about major Senate races has now admitted something fairly important about its methodology: almost none of the polls were real. Wedgewood Polls says all but one of its surveys were fabricated as part of what it describes as a “social experiment.”
The confession came after election-data site FiftyPlusOne started digging into Wedgewood’s raw data and found patterns that did not make sense. When the site contacted the firm for an explanation, Wedgewood’s reported response began with a two-word phrase that is not generally included in professional polling methodology.
Minutes later, the firm publicly admitted the data had been fabricated. It claimed the project was designed to test whether a completely made-up polling organization could gain credibility with aggregators, journalists and political professionals.
The uncomfortable answer appears to be yes. Wedgewood’s numbers were picked up by major polling databases and media outlets, including systems maintained by the New York Times, FiftyPlusOne and other election trackers.
FiftyPlusOne says it had initially vetted the organization in ways that seemed reasonable. Wedgewood supplied names of people supposedly behind the firm, described a plausible methodology and even provided raw survey files that looked convincing enough until analysts began comparing the underlying identifiers with real data from the survey platform it claimed to use.
That is where the story went from sketchy to mathematically impossible. FiftyPlusOne found transaction IDs and respondent patterns that did not behave like authentic Pollfish survey data, including identifiers that appeared to come from months earlier than the dates Wedgewood claimed its interviews occurred.
The analysts then tried to replicate one Wedgewood poll themselves using the actual platform. The deeper they looked, the more the supposedly professional polling operation started to resemble a spreadsheet wearing a fake mustache.
Wedgewood eventually said that was the point. It argued that the exercise exposed how little scrutiny new polling firms face and how quickly political media will amplify data that looks polished, especially when AI tools can create professional websites, PDFs, cross-tabs and methodology documents in minutes.
That explanation has not impressed everyone. FiftyPlusOne says Wedgewood collected money from donors to conduct crowd-funded polls and even supplied fraudulent respondent-level data, making the “experiment” considerably more consequential than posting fake numbers on an anonymous social-media account.
Real campaigns were affected too. Michigan state Sen. Mallory McMorrow says bogus polling hurt her Senate campaign, discouraged supporters and contributed to an environment where reporters questioned whether she should remain in the race.
The scandal also lands in an election cycle already struggling with fake polling operations. Another outfit earlier this year admitted fabricating results as a supposed experiment, meaning the industry has now encountered multiple people apparently deciding the best way to improve trust in polling is to make up polls.
That is a little like improving airport security by sneaking through the terminal with a fake bomb and then acting surprised when everyone gets angry. Wedgewood says it proved the system is vulnerable, but political professionals are left dealing with the fact that fake data was mixed into a very real election.
The larger problem is that polling depends heavily on trust because outsiders cannot independently reproduce every survey. Aggregators can check methodology, sample design and past performance, but a sufficiently determined actor can now generate the paperwork of legitimacy far more easily than before.
AI makes that problem worse because fake cross-tabs, professional-looking reports and plausible websites are cheap to manufacture. The scandal is not proof that polling itself is fake; it is proof that looking like a pollster has become dramatically easier.
Wedgewood wanted to know whether an invented polling company could be treated as real. It got its answer, and unfortunately so did every campaign that spent months reacting to numbers that never existed.
