Become a Votebeat sponsor

The SAVE system isn’t reliable, but it still might be useful. One Arizona county’s experience shows how.

Yavapai County used the federal tool to expand ballot access, find dead voters, and identify a small number of noncitizens.

A close up image of a hand reaching for a pile of stickers that read "I Voted".
"I Voted" stickers are arrayed in the polling station serving wards 194 and 195 in Milwaukee, Wisconsin on Aug. 11, 2026. President Donald Trump’s administration has repurposed a system long used to verify immigrants' eligibility for benefits into a mistake-prone mass voter verification tool. But an election official's experiment in Yavapai County, Arizona, suggests the system, while flawed, may still prove useful in routine voter list maintenance. (Cullen Granzen for Votebeat)

Votebeat is a nonprofit news organization reporting on voting access and election administration across the U.S. Sign up for Votebeat Arizona’s free newsletter here.

Recorder Michelle Burchill was intrigued when she first learned about a newly expanded federal tool for checking voters’ citizenship.

She had helped run elections in bright-red Yavapai County, Arizona, for about six years — and while she entered the field as an elections skeptic, she had gained trust in election systems during her time with the county. Still, Burchill, a Republican, wanted to confidently tell her voters, neighbors, and friends that her county’s voter roll was free of any noncitizen voters.

So, earlier this year, she began running her county’s voter list through the expanded tool, called Systematic Alien Verification for Entitlements, or SAVE.

That system, maintained by the U.S. Department of Homeland Security, was originally built to check immigrants’ eligibility for public benefits. Under President Donald Trump’s administration, it has been transformed into a mass voter verification tool — and a political lightning rod. Trump and his allies have used its findings to claim that there are thousands of noncitizens registered to vote across the country; Democrats have decried it as a voter suppression tool after it was found to have falsely flagged eligible voters as noncitizens.

Burchill, who spent months running her county’s voter rolls through the system, also saw false positives. But she said the system helped her confirm the citizenship status of hundreds of voters who hadn’t submitted a birth certificate, passport, or other documents required by Arizona law to vote a full ballot.

She also found the tool useful for identifying voters who had died and needed to be removed from the rolls, as well as vulnerabilities in the state’s aging voter registration system. And in the process, she found 30 noncitizens on her voter list, representing about 0.01% of her county’s total voters. She reached that number only after using other methods and databases to double-check the citizenship status of voters that the SAVE system initially flagged as ineligible.

Nationwide, virtually every election official aims to maintain a clean voter roll. The politicization of SAVE — which, after a series of court decisions, is now in legal limbo — has problematized its use for routine, nonpartisan maintenance. But Burchill’s experience shows that the system could be helpful — when used responsibly. Her county is now among the first jurisdictions in the country to demonstrate the system’s broader potential as a voter roll maintenance tool.

“The SAVE database is not the only useful tool on which we should rely for voter registration purposes,” Burchill wrote in an Aug. 7 report summarizing her findings, “but in my view, it has been by far the most valuable.”

Nationally, SAVE has made headlines for false positives

Burchill’s experiment stands in stark contrast to the vast majority of jurisdictions nationwide. States and counties that have run their data through SAVE — which tend to be run by Republicans — have largely used it to confirm their suspicions that the rolls are laden with noncitizens. Most Democratic-run jurisdictions, by contrast, have not even used it at all.

Until last year, the tool had a fairly limited use. It could confirm citizenship for certain voters, but only if election officials had specific immigration and naturalization identification numbers issued by the Department of Homeland Security.

But then, in March 2025, Trump issued a sweeping executive order that required the agency to make federal citizenship data available to state and local election officials and provide them with free access to federal systems for verifying voters’ eligibility.

So, late last year, the Trump administration upgraded the system to allow election officials to upload their voter lists to it to scan for noncitizens. And initially, the system indicated thousands of noncitizens were present on their voter rolls. As of June, the Department of Homeland Security said that states had run 60 million voter records through the system — about a third of all registered voters in the U.S. — and 24,000 were flagged as potential noncitizens.

But as election officials began investigating individual cases, they quickly realized that the tool was returning false positives.

In Missouri, the program wrongly flagged hundreds of voters as noncitizens. Officials didn’t attempt to confirm that information before limiting voters’ ability to cast ballots.

In Texas, the system said more than 2,000 of the state’s roughly 18 million voters were noncitizens. County election officials have since determined some of the flagged voters were citizens, though it isn’t clear exactly how many.

In Arizona, just one other county — Maricopa, the state’s most populous — has said it’s used the tool. There, Recorder Justin Heap, a Republican, announced in February that his office had identified 137 potential noncitizens registered to vote. But Heap didn’t answer questions about whether he vetted the SAVE system’s results. Later, he said that he intended to change the registration status of the flagged voters, though experts questioned whether he’d followed the appropriate steps to notify them before doing so.

Prosecutors continue to investigate the voters referred by Heap. They have yet to issue information about whether the voters are noncitizens and whether they could be charged with illegally voting.

Meanwhile, several other jurisdictions, including counties in Texas, have ended up reinstating some voters who were incorrectly flagged as potential noncitizens by the SAVE system.

How Yavapai County ran its voter rolls through SAVE

Burchill began her experiment with the SAVE tool on a small scale — by examining her county’s roughly 300 federal-only voters.

Arizona requires people to submit documents proving their citizenship when registering to vote — a step beyond most states and the federal government, which only require voters to attest to their citizenship under the penalty of perjury. Arizonans who register with a federal form but don’t submit such documents are allowed to vote in federal contests, but not state ones. As such, the state has thousands of federal-only voters — people who haven’t proven citizenship, but are still allowed to vote in presidential and congressional races.

Burchill wondered whether the SAVE system could confirm citizenship for any of them. If so, they could be registered as full-ballot voters, expanding voting access.

Her idea worked. The tool confirmed the citizenship of 255 federal-only voters — all of whom can now vote full ballots. A handful came back as deceased, and several dozen came back with no determination. No noncitizens were found, backing analyses that have suggested most federal-only voters are students, homeless, or born on tribal lands, where birth certificates were not commonly issued for decades.

Burchill then used SAVE to check a list of 6,781 voters impacted by a longstanding state coding error. The coding issue made it difficult to determine whether those individuals had provided proof of citizenship when they registered.

The results surprised her. Several active voters were flagged as deceased, suggesting that there could be significantly more dead voters on the county’s rolls than previously known. Nationally, many states share death information with one another — but not all of them. That means local election officials aren’t necessarily automatically notified when a voter dies.

Another 10 voters — one of whom had previously cast a ballot — were flagged as noncitizens and ultimately had their registrations cancelled.

Before doing so, though, Burchill took additional steps to try to confirm whether these voters actually were deceased or noncitizens. The group of suspected noncitizens were sent formal notices requesting they submit documents proving their status, as prescribed by state law, and given 35 days to respond. Burchill, keeping in mind SAVE’s spotty track record, also personally called, emailed, and mailed letters to those flagged by the tool.

Finally, Burchill decided to run her county’s full voter roll through SAVE. The system confirmed citizenship for the vast majority of voters. It identified 39 as noncitizens, but through further vetting, Burchill said she was able to track down citizenship documents for 28 of them. The other 11 were removed from the voter rolls after the requisite notification and waiting period.

In the process of running the checks, Burchill also discovered nine other noncitizen voters — and a potential glitch in the state’s systems.

In total, she used SAVE to confirm the citizenship of 193,114 voters — 99.8% of those on her county’s roll. Burchill ultimately identified and cancelled the registrations of 30 individuals flagged as noncitizens, four of whom had previously voted.

The takeaways

Burchill’s report concluded the SAVE system, while flawed, can be used to constructively clean voter rolls and build confidence in elections, provided that its data is vetted carefully.

Hers is the most detailed, but not only, example.

Officials in Florida used the tool to help investigate specific voters after receiving information that they might be noncitizens. In a Jan. 15 report, they said they had looked into more than 835 voters and, after vetting them with the SAVE system and other tools, found that about 200 were ineligible to vote.

In Utah, Lt. Gov. Deidre Henderson, a Republican, wrote in a May 27 report that while SAVE could not reliably determine whether a voter was not a citizen without crosschecks, it reliably identified people who were citizens. She said the tool helped “identify and fix potential vulnerabilities in our voter registration processes while also protecting the rights of eligible voters.”

Burchill said her experiment also raised new questions about Arizona’s election laws and systems. For instance, some voters were flagged by the SAVE system as having no record or determination of citizenship. She was able to confirm most of those voters’ citizenship statuses in the Arizona Voter Information Database, the state’s voter registration system — but not all. That means some remain in limbo.

“These voters may very well be citizens, but we don’t have proof of either status,” Burchill wrote, adding that the voters would remain on the rolls under existing Arizona laws.

Burchill said the experiment also raised potential concerns with AVID that state and local officials are still working to fully understand. While attempting to confirm noncitizens flagged by the SAVE system, she discovered that several had seen their citizenship status change in motor vehicle records — without any notification to election officials.

Calli Jones, a spokesperson for the Arizona Secretary of State’s Office, which administers AVID, said officials were looking into the changes described by Burchill. Jones added that her office “does not register voters, nor do we change voter information,” calling the issue “an outstanding question for our office as well.”

And Burchill questioned what election officials should do if they become aware of a potential noncitizen on the rolls within 90 days of an election. Under federal law, officials are barred from conducting systematic removals of ineligible voters during that period.

“This topic has not been addressed in the past but is now something that needs to be addressed in law,” Burchill wrote, adding that the SAVE system created an “uncomfortable” tension with the mandated pause.

SAVE is currently mired in lawsuits. In June, a federal judge blocked the use of the expanded tool, ruling that it violated citizens’ privacy and voting rights. That ruling prompted the Department of Homeland Security to pause certain features in the SAVE system to comply. Last month, a different federal judge ordered the agency to restore access to the revamped system for four states — Florida, Ohio, Iowa, and Indiana — due to an earlier settlement agreement.

Burchill said her use of the SAVE system didn’t infringe on any voters’ privacy. The only information she ran through SAVE was voters’ names, dates of birth, and the last four digits of their Social Security numbers. That information, she said, is already on record in the system.

“I’ve never understood why people keep saying there’s privacy concerns when we are just matching our info to what the database has,” she told Votebeat in a text message. “The whole point is for them to respond and say, ‘Yes, the information is correct and matches.’”

She said she thought it was unfortunate that the courts had limited SAVE’s use and believes the tool can still be valuable.

“I am more confident in our voter roll today than ever before, and that has always been my goal,” she wrote in her report.

Sasha Hupka is a reporter for Votebeat based in Arizona. Contact Sasha at shupka@votebeat.org.

The Latest

Thomas Carroll represented Fulton County and faced a harsh rebuke from the state Supreme Court.

With less than two months until the midterms, the messy court battle is nearing a conclusion.

The charges come as the Trump administration attempts to crack down on voter fraud.

The Michigan Supreme Court said it did not have the power to force canvassers to put the proposed constitutional amendment on the ballot.

Roughly 150,000 voters, most in Maricopa County, received letters asking if they’d moved. Here’s why.

The Election Truth Alliance cited a professor’s statistical analysis to buttress its legal claims, but experts say his model can’t accurately detect fraud in American elections.