President Trump is pushing hard to require documentary proof of citizenship to register to vote, through the SAVE Act in Congress and an executive order of his own. The case for it rests on a claim of widespread fraud. This is a sourced look at two questions that decide whether the policy makes sense: how common is the fraud it targets, and how many eligible American citizens would it shut out. The numbers point the same direction, and it is not the one he claims.
Election integrity is a real value, and wanting to be sure only citizens vote is not a fringe position. The disagreement here is not about the goal. It is about whether this policy fits the problem, and what it would cost the people it sweeps up. So every load-bearing number is attributed to a named source, with these labels:
Documented on the official or court record. Estimate a strong figure from a survey or model. Contested credible sources disagree.
The page states the strongest version of the other side out loud, in its own section near the end. Two quick definitions first, because they are constantly confused: voter ID means showing identification to cast a ballot, mostly under state law. Proof of citizenship, what the SAVE Act and the executive order demand, means producing a passport or birth certificate to register in the first place. They are not the same thing, and the difference turns out to matter a lot.
It would be easy to treat this as one bill among many. The president does not. He has made documentary proof of citizenship a singular demand, and when Congress would not deliver it, he began holding the rest of his own party's agenda hostage to it.
In March 2026 he said he would make no deal with Democrats on anything until they passed it. In April, when Senate Republicans hesitated, he demanded they abolish the filibuster to force it through:
"Not passing the SAVE AMERICA ACT will lead to the worst results for a political party in the HISTORY of the United States Senate. An Unrecoverable Death Wish! Likewise, the FILIBUSTER, TERMINATE IT NOW!"
Donald Trump, Truth Social, April 25, 2026. (Democracy Docket)
It did not work. The SAVE Act failed in the Senate on June 4, 2026, short of the 60 votes the filibuster requires. An earlier attempt to attach it to another bill had failed 48 to 50, with four Republicans (Murkowski, McConnell, Collins, and Tillis) joining every Democrat against it. (NPR; Democracy Docket) Documented
So the hostage-taking escalated. On June 17 he refused to renew a major federal surveillance law unless the SAVE Act rode along with it. And on June 24, 2026, about an hour before he was due at the Capitol to sign a bipartisan housing bill into law, he called the ceremony off:
"Today's Housing News Conference and Signing is hereby cancelled until such time as we pass the desperately needed SAVE AMERICA ACT, which I consider to be a National Emergency."
Donald Trump, Truth Social, June 24, 2026. (NBC News; The Hill on the FISA threat)
Three of his own side's priorities, a surveillance reauthorization, a housing bill, and a Senate majority's patience, spent in a matter of weeks to try to force through a citizenship-document requirement. Documented
The legislative route was only half of it. On March 25, 2025, Trump signed Executive Order 14248, directing the Election Assistance Commission to demand documentary proof of citizenship on the federal voter registration form, with a document list even narrower than the bill's. Two federal courts have since permanently blocked its core: Judge Colleen Kollar-Kotelly in October 2025 and Judge Denise Casper on June 24, 2026, the same day as the housing cancellation. Both held that the Constitution gives control of federal elections to the states and to Congress, not the president. (Brennan Center; Votebeat) Documented
Blocked in Congress and in court, the administration turned to leverage. The Justice Department demanded full voter rolls, including driver's license and Social Security numbers, from nearly every state, and sued dozens that refused, though courts have dismissed many of those suits. Federal officials pressed states to run their rolls through a citizenship database that has wrongly flagged some actual citizens. And by late June 2026, the administration was weighing whether to withhold up to 20% of Homeland Security grant money from states that would not comply. (Brennan Center tracker; NOTUS) Documented
This is the tell. A policy aimed at a problem this small does not usually command this much force. The intensity is itself the argument that something other than the evidence is driving it. The case for the bill still rests entirely on the claim that fraud is widespread, which is where any honest look has to go next.
If noncitizen voting and voter impersonation were common, the case would make itself. Decades of investigation, including by people determined to find fraud, keep arriving at the same answer: it is vanishingly rare.
31 in 1 billion
Credible incidents of in-person voter impersonation, the kind a photo ID would stop, that a nationwide investigation found across more than one billion ballots cast from 2000 to 2014. About one in every 32 million votes. (Justin Levitt, Loyola Law School) Documented
0.0001%
The share of votes that election officials flagged as suspected noncitizen votes in the 2016 jurisdictions with the most immigrants: about 30 out of 23.5 million. Forty of 42 jurisdictions reported none at all. (Brennan Center) Documented
Even the Heritage Foundation, a conservative group that built its election-fraud database to make the case for enforcement, has cataloged only about 10 in-person impersonation cases and roughly 41 noncitizen-voting cases across several decades in which billions of votes were cast. The database is meant to alarm. Its own size is the rebuttal. (Brennan Center analysis of the Heritage database) Documented
Georgia is the test case supporters cite most, so it is worth following the numbers carefully, because they are usually quoted wrong. A 25-year audit flagged 1,634 noncitizens who had attempted to register. Here is what actually happened to them:
A separate 2024 audit of Georgia's active rolls found about 20 noncitizens among roughly 8.2 million registered voters, a rate of 0.0002%. (Georgia Secretary of State; ABC News) Documented
And this is not an unguarded system. Noncitizen voting is already a federal crime punishable by prison, fines, and deportation, and every person who registers must already swear under penalty of perjury that they are a citizen. (18 U.S.C. 611; Dept. of Justice, NVRA) When Trump's own 2017 commission on voter fraud went looking for evidence, it disbanded within a year without producing any. (NPR) Documented
Here is the other half of the ledger, and it is far larger. The documents the SAVE Act would require are exactly the ones millions of citizens do not keep at hand.
21.3 million
Voting-age US citizens, about 9%, who do not have ready access to proof of citizenship (a passport, birth certificate, or naturalization papers). At least 3.8 million have no such document at all. This is a 2023 survey estimate, margin of error about 2.6 points. (Brennan Center / University of Maryland / VoteRiders) Estimate
Not all 69 million, and it is worth being precise, because the honest number is the one that persuades. A passport satisfies the requirement on its own, in whatever name it carries, so a woman who has one is fine despite the mismatch. The people actually at risk are the women who changed their name and do not have a passport (or other current-name document, like naturalization papers). For them a birth certificate alone would not match, and they would have to hunt down marriage or court records to bridge the gap, or be turned away while they do.
It is not hypothetical. Arizona alone reports that more than 1.5 million married women there have changed their surname, and the state demands an exact name match when a birth certificate is used. In New Hampshire's 2025 elections, a Derry woman was sent home for her birth certificate, then had to return again with her marriage certificate when her proof still did not match her license. (Arizona Mirror; New Hampshire Bulletin) Documented
Put the two halves together. On one side, a problem measured in dozens of cases. On the other, a requirement that tens of millions of citizens would struggle to meet. That is the trade the policy actually asks for.
This is not hypothetical. Proof-of-citizenship registration has been tried at the state level, and courts have documented what it does.
Under Secretary of State Kris Kobach, Kansas required documentary proof of citizenship to register from 2013 to 2018. When he defended the law in federal court, the record was lopsided:
A Harvard expert testified the noncitizen registration rate was statistically indistinguishable from zero. The judge struck the law down, and the 10th Circuit affirmed in 2020. (ProPublica; Fish v. Schwab, 10th Cir.) Documented
Arizona's version produced a two-track system after the Supreme Court ruled in 2013 that the state cannot demand documents on the federal form. Citizens who register without papers become "federal-only" voters, about 34,000 of them as of late 2024. In 2024 a state database error also left roughly 97,000 voters' status in limbo weeks before the election. (Arizona v. Inter Tribal Council; Votebeat) Documented
New Hampshire's new proof-of-citizenship law took effect for 2025 elections. A voting-rights group counted at least 244 voters turned away across the year, a number no official record captures because the state does not track denials. Among them, a 23-year-old military veteran in Derry was sent home to retrieve his birth certificate; others included women whose documents did not match their married names. In May 2026 a federal judge struck down the part of the law that had removed the option to sign a sworn affidavit instead. (New Hampshire Public Radio; New Hampshire Bulletin) Estimate
The pattern repeats: a large number of real citizens blocked or burdened, to catch a number of noncitizens so small that courts have called it indistinguishable from zero.
Most people hear "voter ID" and "proof of citizenship" as one idea. They are different, and the difference is the quiet center of this whole debate.
Voter ID means showing identification to vote. It is popular and common: 36 states ask for some form of ID, and polls put support around 80%, including most Democrats. That popularity is real and worth respecting. (NCSL) Documented
But here is the catch. The ID nearly everyone shows is a driver's license, and a driver's license does not prove citizenship. So voter ID, the popular policy, does nothing about the noncitizen-voting claim. To actually screen for citizenship you need the far higher bar the SAVE Act sets: a passport or birth certificate, documents no voter ID law has ever required. The policy riding on voter ID's popularity is not voter ID at all. Documented
An argument worth making has to survive the best version of the other side. Here it is.
Concede all of it, and the conclusion holds. The case for the policy is strongest as a matter of confidence and principle, and weakest exactly where it counts: on whether the problem is big enough to justify the cost.
Strip it to the arithmetic. The fraud the SAVE Act targets shows up in dozens of cases across decades and billions of votes, is already a felony, and is already screened by a sworn citizenship oath. The cost runs the other way: about 21 million citizens without ready proof of citizenship, half the country without a passport, tens of millions of married women whose names do not match their birth certificates, and a documented history of eligible Americans turned away every time a version of this was tried.
That is not a close call disguised as a hard one. It is a policy that would do very little about the thing it names and a great deal to the people it does not. President Trump has pushed it anyway, with a force usually reserved for genuine emergencies: by executive order until two courts struck it down, by holding his own party's bills hostage in the Senate, and, blocked in both, by leaning on states with their own federal funding. The premise underneath all of it, that fraud is widespread, is the one claim his own commission, the courts, and the audits keep failing to support.
There is a genuine problem in American elections, and it is worth taking seriously. It is that eligible citizens, disproportionately the poor, the elderly, rural voters, voters of color, and married women, can be quietly shut out by rules written for a threat that barely exists. A democracy can survive a few illegal votes far more easily than it can survive turning away the real ones. The numbers do not say be careless about citizenship. They say this is the wrong fix, aimed at the wrong problem, and the people who would pay for it did nothing wrong.
Load-bearing numbers are attributed to named sources: government records, court rulings, nonpartisan researchers, and, where noted, the pro-enforcement Heritage database. Survey-based figures are labeled as estimates.
This page assesses the SAVE Act, related proof-of-citizenship measures, and voter ID using public data from government records, court rulings, and nonpartisan researchers as of mid-2026. Figures labeled estimate come from surveys or models and carry uncertainty. Noncitizen voting is real but rare, and it is already illegal; this page does not claim it never happens, only that the evidence does not match the scale of the response. Legislative and court status can change. Corrections welcome.