← back to the guides
Yes. It is also rare, mostly accidental, and nowhere near big enough to decide an election. North Carolina has the paperwork for all of it.
eleven arguments. eleven answers. no vibes, just the state's own paperwork.
Every session in Raleigh brings another bill to secure the elections. Shorter early voting, more paperwork, new ways to challenge a ballot after it has already been cast. Elections should be secure, no argument there. The lucky part is that North Carolina audits itself and publishes what it finds, so nobody has to guess about any of this.
Fraud here is real and it is tiny. After the 2016 election the state checked every ballot cast, nearly 4.8 million of them, and found 508 people who should not have voted. That is one hundredth of one percent, and almost all of them were on probation and thought their rights were back.
The worst case anybody can name, Bladen County in 2018, was run by a paid operative for the Republican campaign, caught under laws already on the books, and thrown out 5 to 0 by a bipartisan board.
The State Board's own words for it: ineligible voting in North Carolina is "neither rampant nor non-existent." That is the whole answer. Everything under this is the eleven arguments you have probably had to sit through at a family function, with something you can actually say back.
"So you're telling me voter fraud doesn't exist."
No, it exists. North Carolina checked all 4.8 million ballots from 2016 and found 508 people who should not have voted, which is one hundredth of one percent. Most of them were on probation or parole and believed their rights had been restored.
After the 2016 general election, the State Board of Elections took every ballot cast, nearly 4.8 million of them, and ran the whole file against felony conviction records, death records, DMV records, and other states' voter files. Not a sample of precincts or a spot check, the entire file.
441 people with felony convictions who voted while still on probation or parole
41 noncitizens with legal status, green card holders and the like
24 people who voted twice, casting 26 ballots between them
2 cases of someone voting in another person's name
508 voters, 510 ballots, out of 4,769,640. Roughly 0.01 percent. Not one race changed.
Both halves of the Board's summary matter: "The evidence suggests that participation by ineligible voters is neither rampant nor non-existent in North Carolina... Most incidents are isolated and uncoordinated."
The same report "strongly cautions readers not to refer to each of these cases as 'voter fraud,'" because fraud is an intent crime, and most of these people did not know they were breaking the law.
so: real, rare, mostly a misunderstanding.That report is the reason I can be specific instead of arguing from impressions. It rules out the claim that the number is zero and the claim that the number is enormous, and it does both with the same set of pages.
Sources: NC State Board of Elections, Post-Election Audit Report, 2016 General Election (April 21, 2017), pp. 4-6 and Appendix 5. Ballot total confirmed as "nearly 4.8 million" in NCSBE's April 16, 2026 press release.
"Bladen County. 2018. Explain that one."
I do not explain it away. It was real, it was the worst election fraud in modern North Carolina history, and it was run by a paid operative for the Republican campaign. The state refused to certify the race, held a four-day hearing, and threw the election out 5 to 0. Every law he broke was already on the books.
Leslie McCrae Dowless was hired to work absentee ballots in the Ninth Congressional District. His crew collected mail ballots from voters' homes, forged witness signatures, and filled in blanks. Investigators called it a coordinated, unlawful, and substantially resourced absentee ballot scheme operating in Bladen and Robeson counties.
The State Board refused to certify the election, in a race the Republican had apparently won by 905 votes.
It ran an investigation with 142 voter interviews and at least 30 subject and witness interviews, then held a four-day public evidentiary hearing.
On February 21, 2019, it voted 5 to 0, Republicans and Democrats together, to throw out the result and order a new election. Mark Harris conceded during his own testimony that a new election was warranted.
Criminal charges followed. Multiple people pleaded guilty.
Everything Dowless did was already a felony under North Carolina law before he did it, so there was no loophole to close afterward. The existing laws are what caught him, and the existing remedy is what erased his results.
It is also worth being precise about what kind of fraud this was. Nobody walked into a precinct pretending to be someone else. This was an organized operation with a payroll, which is part of why investigators could unwind it: an operation that size leaves records, hires people, and pays them.
the guy who says bladen proves the system is brokenSources: NCSBE, "State Board Unanimously Orders New Election in 9th Congressional District" (posted Feb. 25, 2019; the vote was Thursday, Feb. 21, 2019). Bladen absentee irregularities also appear in the 2016 post-election audit, p. 5, where the Board referred the matter to federal prosecutors. Guilty pleas reported by NBC News, Oct. 2022.
"The state referred 270 felons for prosecution in one year. You call that nothing?"
I call it 270 people who voted while still on probation or parole, most of whom thought they were allowed to. There were 278 referrals in 2025 all together, and the other eight were six double voters, one illegal voting case, and one person who photographed their own ballot. Zero for impersonation, zero for noncitizen voting, zero for absentee fraud.
In 2025 the State Board's Investigations Division took in 432 complaints and made 278 criminal referrals. The published breakdown fits on a napkin, so it is worth putting all of it here rather than only the part that helps me.
Felon voter: 270
Double voter: 6
Illegal voting: 1
Photographing a voted ballot: 1
Voter impersonation: zero
Noncitizen registration or voting: zero
Absentee ballot fraud: zero
That is the complete list and it adds to 278. The Board did receive complaints in the three zero categories that year, six alleging impersonation, eight involving noncitizens, and five alleging absentee fraud. It referred none of them, because they did not hold up.
Almost everything the state catches falls into that single category, which is people who believed they were eligible and were mistaken about it. The categories most people picture when they hear the word fraud produced no referrals at all that year.
There is a reason so many of them are confused. On August 23, 2021 a state court ruled that North Carolinians on probation and parole could register, and eleven days later the state Supreme Court put that on hold. In March 2022 the trial court struck the disqualification down for good, and once the appeals shook out the State Board opened registration on July 27, 2022 to roughly 56,000 people. They registered. They voted that November. On April 28, 2023 the state Supreme Court reversed and took it back.
So the correct answer to "can I vote" changed four times in twenty months, and the last change landed after these people had already cast a ballot. North Carolina is also stricter than most people assume: finishing your prison sentence is not enough here, you have to finish probation, parole, and post-release supervision too.
If the goal is to get that 270 closer to zero, the most direct way there is to tell people in writing, on the day supervision ends, that they can register again. It costs almost nothing to send a letter, and it addresses the one thing that is actually generating these cases. Nothing currently moving in Raleigh does that.
Sources: NCSBE Investigations Division, 2025 Incident and Case Data, the full complaint and referral tables (432 complaints, 278 referrals). Rights-restoration timeline: NCSBE, statement on the September 2021 stay; NCSBE, "North Carolinians Serving Felony Sentences, Who Are Not in Jail or Prison, May Register to Vote Starting July 27" (July 26, 2022); Community Success Initiative v. Moore, N.C. Supreme Court, decided April 28, 2023, reversing the trial court judgment.
"Homeland Security found 1,599 noncitizens on our voter rolls. Tell me that's fine."
It is not fine, and the state said so and is working through the cases. Read the number carefully, though. The state's own word is "potential," meaning flagged for review rather than proven ineligible, and these are registrations, not ballots, out of 7.4 million records checked. Nobody has reported how many of the 1,599, if any, actually voted.
In April 2026 the State Board signed an agreement with U.S. Citizenship and Immigration Services and submitted 7,397,734 voter records to the federal SAVE database to check citizenship. It came back with 1,599 potential noncitizen registrations, about two hundredths of one percent of the file.
"Potential." SAVE flagged a possible mismatch. It did not prove anyone was ineligible. That is why the Board adopted rules requiring notice and a chance to produce documents before anyone is removed.
"Registrations." Being on a list is not casting a ballot. The Board has not reported how many of the 1,599, if any, actually voted.
Sam Hayes, the Board's executive director, called the 1,599 "an extremely small percentage" of North Carolina's roughly 7.8 million registered voters, and added that "one ineligible registrant is still one too many." I agree with him on both halves.
"Election officials do not have evidence to suggest noncitizen registering and voting is a widespread problem. However, there are documented cases of noncitizens making their way onto the voter rolls, often by mistake."
That wording comes from the agency conducting the check, in the announcement of the check, under an executive director appointed by a Republican-controlled board. It was not written to help my argument, and I would not be able to quote it if it had been.
check the rolls. i want the rolls checked.Sources: NCSBE, "State Board to Check Voter Rolls to Identify, Remove Non-U.S. Citizens" (April 16, 2026); NCSBE, April 27, 2026 release (7,397,734 records submitted April 17, 2026); Hayes quotes via Carolina Journal, "NC investigating 1,599 potential noncitizen voter registrations" (July 20, 2026).
"Thirty-four thousand dead people on the voter rolls. THIRTY-FOUR THOUSAND."
Dead registrations, which are records in a file, not dead voters. The state is the one that found them, by running its own file against a federal database, and it is removing them. Nobody has produced evidence that a ballot was cast in any of those names.
Same SAVE run, same 7.4 million records. About 34,000 came back matching people who had died. Hayes said plainly: "While we expected to find some cases, this is higher than we anticipated."
The Board also explained how it happened. North Carolina gets death records weekly from the state Department of Health and Human Services, which catches you if you die in North Carolina. It does not catch you if you registered in Winston-Salem, moved to Florida in 2011, and died there. The federal match found that gap, which is what the match was for.
"The identification of deceased individuals on the voter rolls does not necessarily indicate that illegal votes were cast in their names, but it does underscore the importance of regular updates and strong interagency coordination."
Turning one of those records into a vote requires someone to physically show up and claim to be that person, which is a separate thing that the state has also gone looking for. The 2016 audit pulled 19 records where the death date preceded the vote date. Most were not fraud at all. The report's example: the death record was for John Smith Sr. and the voter record was for John Smith Jr.
Out of nearly 4.8 million ballots, two people voted in someone else's name. A widow who mailed her late husband's ballot after forging his signature. A daughter who voted in her deceased mother's name. Both admitted it. From the report: "The suspects in each case indicated that they were motivated out of a desire to carry out their loved one's voting wishes." Both were referred to prosecutors.
Both of those cases came out of a death in the family rather than a scheme, and the state identified both of them. That is the entire body of impersonation the audit turned up in a presidential election.
The cleanup also runs continuously rather than in response to a scandal. In the first three months of 2025 alone, county boards removed 25,119 deceased registrations as part of routine list maintenance required by federal law.
The John Locke Foundation is not on my side of much, but when they wrote up the same 34,000 they noted that officials "will not automatically remove those registrations" and that county officials "will follow procedures to verify the data and provide due process." Nobody in that story claimed a dead person voted, because nobody had evidence one did.
Sources: NCSBE, "State Board Identifies Deceased Individuals on Voter Rolls Through Federal Database Comparison" (April 27, 2026); NCSBE, "NC Counties Conduct Voter List Maintenance as Required by Law" (May 1, 2025); 2016 post-election audit, p. 6; Andy Jackson, John Locke Foundation, "Bring out your dead!" (April 28, 2026).
"How does a noncitizen even get on the rolls? Somebody is letting them."
Usually by being told wrong. The 2016 audit found 41 noncitizens who voted out of nearly 4.8 million ballots, all of them here legally, and investigators sat down and interviewed them. Several had been misinformed at voter registration drives, and in at least one case by a precinct official.
The 2016 audit found 41 noncitizens who voted, all of them here legally, green card holders and similar. Investigators did not stop at the number. They talked to people.
Some were misinformed by individuals conducting voter registration drives.
In at least one documented case, they were misinformed by a local precinct official.
One registrant in her seventies had lived in the United States for more than fifty years and believed she was a citizen because she had been married to one.
None of that describes an organized effort. It describes a woman in her seventies working from a fifty-year-old assumption, a volunteer at a folding table who had the law wrong, and a precinct official who should have known better than either of them.
Deliberate cases do exist and they get prosecuted. The State Board points to a Canadian citizen who recently pleaded guilty in federal court to two counts of falsely certifying citizenship on North Carolina voter registration applications, in 2022 and again in 2024.
Sources: NCSBE, April 16, 2026 press release, which summarizes the 2016 audit findings and the federal guilty plea; 2016 post-election audit, p. 4.
"You only catch the dumb ones. How would you know about the fraud you missed?"
You are right that nobody can prove a negative, and I am not going to pretend otherwise. What I can tell you is that North Carolina's audit started from all 4.8 million ballots rather than from complaints, which is built to find the cases nobody reported, and it still came back with 508.
North Carolina does not wait for someone to file a complaint. The 2016 audit worked the other direction. It began with every ballot and cross-checked the entire file against felony records, death records, DMV records, and other states' voter rolls. That is how you find things nobody reported, and it is why this state can answer the question at all.
The Presidential Advisory Commission on Election Integrity, 2017. Created to find the millions of illegal votes President Trump said existed. Chaired by Vice President Pence. Dissolved eight months later, January 3, 2018. Commissioner Matthew Dunlap sued for its records; the draft report's section on evidence of voter fraud was, in his description, "glaringly empty."
The Associated Press, 2021. Reviewed every potential fraud case in the six 2020 battleground states Trump disputed. Found fewer than 475, against a combined margin of 311,257 votes out of 25.5 million ballots.
Justin Levitt's survey, 2014. Tracked credible voter impersonation from 2000 through 2014 and found 31 incidents out of more than one billion ballots cast.
The claim that the real fraud is the fraud nobody caught cannot be disproven, which is why it survives every audit that gets published. It is also a claim that applies to anything imperfectly measured, from shoplifting to tax evasion, and the usual response to those is not to make the legitimate version of the activity harder.
Sources: NCSBE 2016 post-election audit, methodology at pp. 3-4. PACEI dissolved by executive order Jan. 3, 2018; Dunlap's account reported by NPR (Aug. 4, 2018) and U.S. News (Jan. 10, 2018). AP, "Far too little vote fraud to tip election to Trump, AP finds" (Dec. 14, 2021). Levitt via The Washington Post (Aug. 6, 2014).
"Riggs won by 734 votes. In a race that close, fraud decides it."
You are right that a close race is where it would matter, which is exactly what makes this one useful. Griffin spent six months challenging more than 65,000 ballots with full access to the voter file. Almost all of it came down to missing data fields on registration forms, and the rest to a residency rule the legislature wrote in 2011, which he actually won in state court. Not one of the 65,000 challenges alleged a fake voter, a double vote, or a ballot cast by someone who did not exist.
2024, State Supreme Court. After a statewide machine recount, Allison Riggs led Jefferson Griffin by 734 votes out of more than five and a half million cast. Griffin did not concede. He challenged more than 65,000 ballots and litigated for six months, through state courts and federal courts.
That is about as favorable a set of conditions for finding fraud as anyone is ever going to get: a well-funded legal team with every incentive to look, full access to the voter file, six months to work, and a target of only 734 votes.
About 60,000 voters whose registration records were missing a driver's license number or the last four digits of a Social Security number, a data-entry field on a form.
About 1,500 military and overseas voters who had not sent in a photocopy of their ID, which the rules in place when they voted did not ask them for.
267 overseas voters, a figure later filings put higher, who had never lived in North Carolina but whose parent was last eligible to vote here. A 2011 state law says those voters may vote in North Carolina.
I want to be straight about that third category, because it is the one where Griffin had a real legal argument and he won it. The Court of Appeals held in April 2025 that a person who has never lived here cannot claim North Carolina as a residence, and the state Supreme Court let that ruling stand. Reasonable people land on both sides of that question.
It is still not fraud. Those voters did exactly what North Carolina's own statute told them they could do, on a provision the General Assembly wrote in 2011 and has left on the books ever since. The fight was over whether that law should exist, not over whether anybody lied to get a ballot. Across all three categories, in six months of litigation, not one challenge alleged that a fake person had voted, that anyone voted twice, or that a ballot was cast by someone who did not exist.
A federal court ordered the election certified on May 5, 2025, holding that a state cannot change the rules of an election after the ballots are in. Griffin conceded two days later.
So the premise of the question is correct. Close races are where fraud would matter most, and that is precisely why this one was examined more thoroughly than any other election in state history. What the examination produced was an argument about form fields and a fifteen-year-old residency statute.
six months of looking, 65,000 ballots pulled,Sources: NCSBE, "State Board Will Not Order Full Recount in NC Supreme Court Contest" (Dec. 10, 2024); machine recount result reported by WUNC (Dec. 4, 2024); challenge categories, the April 11, 2025 state Supreme Court order, and the May 5, 2025 federal order reported by Votebeat and State Court Report; concession reported by AP, May 7, 2025. The overseas residency provision is N.C.G.S. § 163-258.2(1)e., enacted in 2011 as part of the Uniform Military and Overseas Voters Act.
"The Heritage Foundation has an entire database of fraud cases. Explain that."
Heritage answers this one themselves, in the introduction to the database: it is "a sampling" of cases and "not comprehensive." It is a collection of prosecutions gathered across decades and all fifty states, so it can tell you fraud happens. It cannot tell you how often, because a collection has no denominator.
"The Election Fraud Map displays a sampling of proven instances of election fraud... While not comprehensive, this map demonstrates vulnerabilities in the election process and the ways in which fraud is committed."
Those are Heritage's own words on their own site, and there is nothing wrong with the project they describe. The database is a set of cases they went out and gathered, spanning decades and every state, to illustrate what election fraud looks like when it occurs. Building that is legitimate work.
The problem is what happens when the pile gets treated as a measurement. To calculate a rate you need to know how many ballots were cast over the same period, using a consistent standard for which cases count. The database has none of that, which is presumably why Heritage does not publish a rate either.
Sources: Heritage Foundation, Election Fraud Map introduction, quoted verbatim from the archived version of electionfraud.heritage.org captured July 16, 2026. Comparison figures from the NCSBE 2016 post-election audit.
"You're a Democrat running for office. Of course you say there's no fraud."
That is a fair suspicion, so start with the number that cuts against me. In the 2016 audit, registered Democrats were over-represented among ineligible voters, 326 of the 508, against a 39.5 percent share of registration. It is in the report, I read it, and I am not going to bury it in a footnote.
The 2016 audit broke every case down by the voter's party registration. The whole table, including the row I would rather skip:
Felony supervision: 292 Democrat, 73 Republican, 72 unaffiliated, 4 Libertarian
Voting in another person's name: 2 Republican, 0 everyone else
All categories: 326 Democrat, 91 Republican, 87 unaffiliated, 4 Libertarian
North Carolina's registration that year was 39.5% Democrat, 30.2% Republican, 29.9% unaffiliated, 0.4% Libertarian.
Run the comparison and registered Democrats come out over-represented in those cases, substantially so in the felony-supervision category. That is in the report, and I am not going to pretend I did not read it or push it into a footnote.
I include the two Republican impersonation cases for the same reason. Two cases cannot tell you anything about a party with two million registrants, and neither figure is worth much as a partisan argument. If I showed you only the rows that helped me, you would be right not to trust the rest of the page.
The State Board's own conclusion, which I think is the correct one: "ineligible voters are not isolated to one political party or any geographical region of the state."
Close to home, since this is District 31: in 2016 Forsyth County had 20 felony-supervision cases and 2 noncitizen cases. Stokes County had 2, both felony supervision.
Sources: NCSBE 2016 post-election audit, Appendix 5 (party breakdown and the 2016 statewide registration baseline in the appendix footnote) and the county-level table; conclusion quoted from p. 6. Hayes quote from NCSBE's April 16, 2026 release.
"Fine. It's small. But what's the harm in being careful?"
None at all, if you mean the rolls, and I would vote for the careful parts of House Bill 958. The rest of the bill is about the calendar rather than the rolls. Cutting primary early voting from 17 days to 10 does not make one person eligible or ineligible. It takes a week away from people who cannot get to the polls on a Tuesday, and a shorter window means fewer of them vote.
"Being careful" already looks like this. In the first three months of 2025, county boards removed nearly 499,000 registrations under the National Voter Registration Act and state law:
419,937 inactive eight years or more · 40,357 moved within the state · 25,119 deceased · 7,017 duplicates · 2,849 felony convictions · 2,784 moved out of state · 514 removed by request · 5 noncitizens · 307 other
That is one quarter of one year, and it is the same machinery that produced the 34,000 dead registrations and the 1,599 citizenship flags. Those numbers exist because the checking is already happening.
House Bill 958 is what is actually in front of the General Assembly. As of August 4, 2026, it is not law: the Senate passed its version on July 28 and the House has a concurrence vote calendared for August 4. That vote may already have happened by the time you are reading this, so check the bill's page before you quote me on where it stands.
Parts of it are genuinely fine by me, and it would be strange if they were not, given that this entire page is built out of the results of checking. It adds a citizenship attestation to the registration process. It also gives the State Auditor authority to run post-election audits of counties, and that section is built with more care than most of what comes out of this legislature: counties are selected at random, no audit can begin until after the election is certified, no finding can be used to challenge a result, every county gets audited at least once every six years, the annual reports have to be published and stay published, the county under review gets to read the report and attach its own response, and the audit manual is written jointly with the State Board and frozen for the ninety days before an election.
I would vote for that section. The asterisk is not about the audits, it is about who holds them. Since 2025 the State Auditor also appoints the State Board of Elections and the chairs of all 100 county boards, so this bill would put one office in charge of picking the people who run elections and then grading the counties where they run them. Sen. Val Applewhite said it on the floor better than I can: "He audits his own people. He grades his own homework." That is a structural objection and not a personal one. I would say the same with a Democrat in the chair, and the law that moved those appointments is still being fought over in court, so where the power sits may not be settled yet. Put the audit authority somewhere that does not also appoint the board, and I am all the way in.
Early voting for primaries, second primaries, and runoffs shrinks from 17 days to 10. The period would start on the second Thursday before the election instead of the third.
Special and municipal elections get capped at 10 consecutive days of early voting.
Ballots already cast during early voting become challengeable for the first time. Any registered voter in your county could file a challenge against a ballot you already cast, within three business days.
General elections keep all 17 days.
Those first two are calendar changes rather than roll changes, and they belong in a separate conversation from everything else on this page. Early voting exists because a lot of people cannot vote on a Tuesday. Take a week out of it and the same people are still eligible, still registered, and still counted if they make it. Some of them will not make it, and the ones most likely not to are shift workers, parents managing school pickup, and anyone whose one reliable free day is a Saturday.
Two amendments came up during the July 28 Senate debate that would have made the careful part work better. One would have joined North Carolina to ERIC, the multistate compact that finds people who moved or died by comparing voter records across state lines, which is a more reliable instrument than the federal databases the state is currently leaning on. The other would have required that a voter flagged as deceased get notified before being dropped from the rolls, since name collisions are a documented weakness of the SAVE system and the person flagged is sometimes alive. Both were voted down, along with about ten other amendments.
So be as careful as you want with the rolls. Run the database checks, audit the counties, prosecute Dowless and anyone like him, and I will help. I would just rather not pay for that care with somebody else's ballot.
Sources: NCSBE, "NC Counties Conduct Voter List Maintenance as Required by Law" (May 1, 2025), removals under 52 U.S.C. § 20507 and N.C.G.S. § 163-82.14. Bill provisions read directly from H958, Eighth Edition Engrossed, Senate committee substitute adopted and engrossed July 28, 2026; House concurrence calendared for August 4, 2026. Provisions cited take effect January 1, 2027. Audit provisions quoted from new G.S. 147-64.6Q, Section 3.3(a) of that edition. Applewhite quotation and Sen. Warren Daniel's response reported by Carolina Journal (July 29, 2026). Rejected amendments reported by Lynn Bonner, NC Newsline (July 28, 2026). The State Auditor's appointment power over the State Board dates to S.L. 2024-57 (Senate Bill 382) and remains in litigation.
Eleven arguments, eleven answers, one screenshot. Long-press it, text it to whoever needs it, and come back here when somebody wants to see where the numbers came from.
Accurate rolls and accessible voting are not in tension, and North Carolina is currently demonstrating that in public with published numbers. The state ran a federal citizenship check, found gaps in its own file, and began working through them under rules that give people notice and a chance to respond. That is what being careful looks like when it is done well.
The disagreement in Raleigh is less about whether to be careful than about who absorbs the cost of it. A policy that makes the rolls more accurate without costing an eligible voter anything has my support. A policy that costs an eligible voter a day of work needs a better justification than the word security.
I would also like the rest of the session back. Time spent shaving a week off the primary calendar is time not spent on schools that remain underfunded, or on the law that shifted $24.8 million in power costs onto residential bills, or on what a night in the hospital costs in this state. Those are the pressures people in District 31 are actually under right now, and the early voting calendar has never once come up when someone tells me what is making their life harder.