The legacy press has an entire vocabulary for noncitizen voting, and every word carries the same instruction: Stop looking. U.S. News and PBS call it “extremely rare.” Time says “very, very rare.” NPR quoted an expert calling it “infinitesimally rare.”
But Fairfax County, Virginia tells a different story — one the media won’t touch.
Fairfax County referred hundreds of potential noncitizen voters for prosecution over seventeen months. Not one was charged. Then officials abandoned the policy as “too burdensome.”
Between September 2024 and February 2026, Fairfax County’s electoral board required the registrar to refer suspected noncitizens to prosecutors. In a single batch last October, 71 individuals were flagged. An earlier summer batch sent 78 names forward. Over seventeen months, the cumulative total ran well into the hundreds.
Yet on Feb. 19, 2026, the board voted to kill the policy. The stated reason: not a single prosecution. Compliance consumed “significant” staff time at the elections office and prosecutors’ offices. Officials concluded no prosecutor would find checking a DMV box met the beyond-a-reasonable-doubt standard.
Read that again. Hundreds of referrals. Zero prosecutions. Policy scrapped.
And here’s what matters for the “rare” claim: Because none were prosecuted, not one will ever appear in databases journalists cite to prove noncitizen voting doesn’t happen. The cases were found. They were referred. They were dropped. Their disappearance from the record is then used as proof they didn’t exist.
The context is worse. Fairfax County canceled 985 registrations on citizenship grounds between January 2022 and July 2024 alone. County Republicans count 1,912 cancellations over four years — including 538 in 2025, more than one per day. Virginia removed 6,303 noncitizen registrations statewide in the same period.
Crucially, Virginia does not verify citizenship at registration — federal law forbids requiring documentary proof on the federal form. These registrations surfaced only when registrants disclosed their own status, typically on a jury questionnaire or DMV transaction. Nobody went looking. The system caught only people who raised their hands.
Every fraudulent vote cast invalidates the vote of an eligible citizen. A noncitizen who votes doesn’t merely add a vote to one side — he subtracts one from a citizen who did everything right.
How many disenfranchised citizens is acceptable? One hundred? A thousand? We don’t apply this reasoning anywhere else. Nobody argues ballot-box stuffing is tolerable at low volume. We don’t tell mugging victims street crime is rare.
Rarity is irrelevant to the margin. Recent congressional races have been decided by a few hundred votes. Virginia’s House of Delegates control came down to a tied district settled by drawing a name from a bowl. In races like these, “infinitesimally rare” fraud is more than sufficient to determine who governs.
The Heritage Foundation’s Election Fraud Database documents 1,619 proven instances of election fraud — cases ending in conviction, civil penalty, judicial finding, or overturned election. The database doesn’t include potential fraud officials discover but never prosecute. It obviously can’t include fraud that goes undetected.
This is the circularity at the heart of the “fraud is rare” consensus. The statistic measures prosecutions. It’s offered as proof about incidence. But you can’t establish the rate of a crime by counting convictions when authorities decline to prosecute — especially when the system is designed so the crime is rarely discovered at all.
Similar findings have emerged nationwide:
Ohio: Secretary of State Frank LaRose identified 597 noncitizens who registered to vote in August 2024. Of those, 138 cast ballots. He referred them to the attorney general.
Texas: Gov. Greg Abbott announced more than 6,500 noncitizens removed from rolls in the three years before the 2024 election. After that election, the secretary of state referred 33 potential noncitizen voters for criminal investigation.
Alabama: Secretary of State Wes Allen identified 3,251 registered voters who had been issued noncitizen identification numbers. He directed county boards to move them to inactive status pending verification.
Tennessee: Election officials sent letters to more than 14,000 registered voters identified through state records, asking them to verify U.S. citizenship.
Louisiana: Gov. Jeff Landry ordered the Office of Motor Vehicles to compile a list of people issued temporary licenses or ID cards for cross-checking against voter files. Louisiana’s OMV had issued more than 40,000 temporary IDs to noncitizens and non-permanent residents.
North Carolina: DHS identified 1,599 potential noncitizens on voter rolls in June 2026. Federal prosecutors indicted and sentenced a Canadian citizen to federal prison for falsely claiming U.S. citizenship and voting in both 2022 and 2024 elections.
Virginia: During the fourteen months from May 2023 to June 2024, 1,973 unique voter records were marked for removal with the reason “Determined Non-Citizen.” Of those, 399 had cast one or more ballots, according to the Electoral Process Education Corp.
Officials in these states will say the numbers are preliminary, that some flagged individuals turn out to be naturalized citizens, and that registration isn’t the same as voting. All true. But note what those caveats concede: Nobody actually knows. Which is precisely the point.
The defensible sentence is not “Noncitizen voting is extremely rare.” The defensible sentence is: We do not systematically verify citizenship at registration, we detect ineligible registrations mostly by accident, and when we do detect them, we usually decline to prosecute.
That is a different claim entirely. The first implies a settled empirical finding. The second describes a system that has arranged, more or less deliberately, not to know the answer — and then treats its own ignorance as reassurance.
Fairfax County spent seventeen months finding cases and referring them to prosecutors who did nothing. Then it stopped looking.
Whatever else that is, it’s not evidence there was nothing to find.
