Two provisional ballots in Texas were rejected before they could enter the count. That detail appears in the federal prosecutors’ own account of a September election-enforcement operation. It matters because new cases like these are being presented alongside a much larger political claim: that Donald Trump was the rightful winner of the 2020 presidential election. An alleged attempt to cast an ineligible ballot, a counted illegal vote and evidence sufficient to change an election outcome are three different things. [6] [9]
The immediate source of the two items reviewed here is Trump’s own Truth Social account. On 26 September he shared two Just the News reports: one about a crackdown involving hundreds of arrests and indictments, another about four people charged in Pennsylvania. Those posts amplify news headlines. Neither of the two posts itself says that the newly accused people have been convicted or that their cases overturn 2020. It would be inaccurate to attribute those additional words to them. [1] [2] [3] [4]
There is, however, a direct and documented connection to Trump’s claim of victory. In a separate Truth Social post on 18 August, he cited a new Census Bureau analysis of the 2020 electorate and wrote: “I WON THE ELECTION!” He then called for the SAVE America Act. That explicit statement, rather than a motive inferred from a headline, is the claim we can test. [9] [10]
The four Pennsylvania prosecutions are real. The Justice Department announced them on 24 September. Three defendants face a total of four counts concerning alleged votes in the 2022 or 2024 general elections. The fourth faces a registration-application charge concerning February 2026; prosecutors also allege an earlier vote, but did not announce a voting charge against that defendant in this release. The department describes two defendants as lawful permanent residents and two as unlawfully present. Immigration status and US citizenship are different questions. These are criminal accusations, not findings of guilt, and the defendants are presumed innocent unless proved guilty. None of the charged acts described in this announcement concerns voting in the 2020 presidential election. [5] [21]
The national headline requires a different check. Just the News cites 67 people indicted or convicted in a period described as recent weeks, alongside more than 160 Homeland Security arrests. Those categories do not supply a reconciled count of distinct people convicted of casting illegal votes. Some people may appear in more than one enforcement total; the report does not provide the person-level information needed to determine the overlap. That is a limitation, not proof that a specific number was double-counted. [3]
A Justice Department spokesperson told the Associated Press that its figure of 70 defendants covered unlawful voting or registration charges, while many of the 160 HSI arrests arising from voter-fraud leads involved related offences, including false citizenship claims or false statements. The figures covered January 2025 through September 2026. An arrest count is therefore neither a conviction count nor a count of illegal ballots added to an election result. Nor can arrests divided by the electorate establish the true fraud rate: detection, selection and unresolved cases all affect that calculation. [7]
This does not mean that illegal voting never happens. In January 2026, federal prosecutors in Michigan reported a guilty plea by a man who had illegally voted in 2020 while using another identity. That is substantially different evidence from a new charge or a database flag. It establishes an individual offence; it does not establish an operation large enough, or distributed in the necessary states, to reverse the presidential result. [8] [11]
There are also genuine administrative failures. New Jersey’s governor reported in July that a software problem had affected about 6,600 automatic registrations during 2023–24, with fewer than 400 people in the flagged group recorded as having voted. A follow-up announced removal proceedings for more than 4,500 registrations and further county review. These official accounts warrant scrutiny. They do not establish deliberate fraud by every affected person, and events in 2023–24 cannot change the votes cast in 2020. A sound fact-check must acknowledge such failures instead of treating every concern about election administration as invented. [19] [20]
Trump’s strongest apparent new evidence about 2020 is the Census report. Published in August 2026 as a beginning analysis, it classifies more than 24,000 records as noncitizen voters after matching a commercial voter file to federal administrative data. It is a real government report, not a fabricated citation. But a classification produced by matching databases is not the same as individual verification that a particular person unlawfully cast a counted ballot. [10]
The version examined for this investigation does describe safeguards. It uses high-confidence identity matches and screens for citizenship evidence, including naturalization records. It says approximately 63,800 people with earlier green-card records who naturalized before the election were excluded from the noncitizen count. We therefore do not adopt the sweeping criticism that the analysis simply ignores naturalization. [10]
The remaining question is how often its data or classifications are wrong. The report does not publish an empirical false-positive rate for this particular group or a case-by-case legal determination. Election-law scholar Justin Levitt identifies potential errors in the commercial voter file, federal records and the matching process. His criticism supplies reasons to test the result; it does not prove that every flagged record is mistaken. Conversely, a high-confidence identity match does not by itself establish that every citizenship record is complete and correct for Election Day. The defensible conclusion is that the report produces findings requiring validation, not 24,000 individually proved crimes. [10] [12]
There is a second, independent problem with Trump’s claim: the arithmetic. The report’s rounded classifications are about 1,100 in Arizona, 400 in Georgia and 500 in Wisconsin. Biden’s respective official margins, calculated from Federal Election Commission totals, were 10,457, 11,779 and 20,682. The same comparison yields 1,000 classifications against an 80,555-vote margin in Pennsylvania, 600 against 154,188 in Michigan, and 250 against 33,596 in Nevada. [10] [11]
Even if every flagged record in each of those six states represented an illegal, counted vote for Biden, removing those votes would not reverse the result in any of them. That deliberately generous assumption is not a claim about how these people voted: the report does not show their candidate choices. The analysis is also incomplete, so these figures are not a ceiling on anything later investigation could find. But an assertion that unexamined records will change the outcome remains a prediction requiring evidence. The findings Trump cited do not demonstrate that he won. The certified electoral result remains Biden 306, Trump 232. [9] [10] [11]
The litigation record reinforces that conclusion, but needs accurate wording. It is wrong to say every judge proved that no fraud of any kind occurred. Some cases ended because of standing or other procedural problems; some were withdrawn; others received substantive consideration. A review by the conservative authors of Lost, Not Stolen classified 64 challenges brought by Trump and his supporters: 20 ended before consideration of the merits, 14 were withdrawn and 30 were addressed on the merits. Their account identifies one limited Pennsylvania success that did not affect the winner. This is an attributed review, not a fresh audit by us of every docket. [15]
The primary court records illustrate the distinction. In Texas v. Pennsylvania, the Supreme Court rejected the case for lack of standing on 11 December 2020. That order did not adjudicate every fraud allegation. In a separate Pennsylvania appeal on 27 November, the Third Circuit rejected the campaign’s challenge and explained the inadequacy of its allegations and requested relief. It also noted that the campaign was not pleading a fraud case. Neither decision supports the claim that Trump was judicially recognized as the election’s winner. [13] [14]
Election checks are another kind of evidence. Georgia’s statewide hand count and machine recount confirmed the winner. A separate Cobb County signature audit examined 15,118 envelopes and reported no fraud in that sample. A sample is not proof that no offence existed anywhere, but it is a concrete check on allegations, not merely an appeal to authority. Taken together, the certified totals, recounts and failed outcome-changing challenges do not support Trump’s claim that fraud deprived him of victory. [11] [15] [16]
The accurate verdict has two parts. Reporting that four people were charged is supported. Treating charges, arrests and preliminary database matches as proof that Trump won in 2020 is not. His explicit claim of victory contradicts the certified result, and the new evidence he cites does not establish a reason to overturn it. Investigating actual offences and correcting administrative failures are compatible with rejecting that unsupported leap. [5] [9] [10] [11]