Key Points

• AI tools speed up signature review process in candidate legal challenges

• Software used in 2024 election proved faulty, causing lawsuit dismissals 

• Attorneys continue to push human review as crucial to ensuring accuracy 

A candidate’s place on the ballot is a numbers game. And now, developments in artificial intelligence software created to review candidate petitions streamline the play. 

Over the past few election cycles, candidate legal challenges — the process of suing to knock someone from the ballot for failure to secure enough signatures — have become increasingly aided by automated petition review software. 

But even though the software can now run line-by-line through signatures, having licensed attorneys verify the results remains crucial, especially after a failure of a nascent petition review software last cycle led to the dismissal of four lawsuits. 

Still, law firms plan to deploy software again this cycle. Election attorneys pioneering the technology, and those opting for the old-fashioned way, think an all-out replacement for human review is still way off. 

“Maybe at some time in our lifetimes, AI will be a substitute for basic human judgment,” election attorney Kory Langhofer said. “But it’s not there yet.” 

It’s candidate filing season, leaving those running for statewide and legislative office to turn in nomination petitions bearing enough valid signatures to make it on the ballot. 

And immediately after the Secretary of State’s Office doors shut and lock at 5 p.m. on the filing deadline of March 23, the ten-day window for candidate legal challenges opens. 

No government official initially takes on the task of combing through candidates’ petitions. Instead, checks on signatures are made solely by private litigants. 

“Let’s say you needed 5,000 signatures. You could write Mickey Mouse down 5,000 times, and unless someone challenges you, you’re going to be on the ballot,” election attorney Tim La Sota said. “We rely on private litigants to enforce these signature requirements, otherwise, there’d be no incentive to get actual good signatures.” 

A “good signature” comes from a qualified elector registered to vote in the candidate’s jurisdiction and either registered with the same political party as the candidate or as an unaffiliated voter. 

Electors can only sign one petition for a single office, unless, like in House races, more than one candidate is running. And beyond the signature itself, a signer must jot down a legible printed name, address and date. 

Under state law, any elector is entitled to litigate a candidate’s eligibility over signatures, and many do. 

In 2024, 21 candidates faced legal challenges. Of the 21 candidates, 11 withdrew, two were disqualified by court order, six stayed on the ballot after cases were dismissed, and two kept their candidacy after prevailing in court. 

Candidate nomination challenges require a line-by-line review of petition sheets. For both the plaintiffs and the defense, that often means close reviews and fights over the many nuances of election season. 

La Sota, a Republican election attorney, said he still sues and defends candidate petition challenges the “old fashioned way” by having reviewers go line-by-line, or audit a sample of signatures and dig deeper in the event of a high error rate. 

But in recent cycles, some election attorneys are turning away from old habits, opting instead to pursue automated review processes.

Langhofer, a Republican election attorney at Statecraft, co-founded Signafide in 2018, a machine learning software trained to go line-by-line to identify petition defects and invalid signatures. 

Signafide uses AI to complete its own assessments and to speed the process for living, breathing signature review personnel. But, because AI can be error prone, Langhofer said at least two human reviewers always check the resulting reports before a signature is flagged.

In 2024, a competitor to Signafide emerged in Veracity Petitions, LLC. The software, spearheaded by election attorneys Brett Johnson and Eric Spencer at Snell and Wilmer, rolled out in 2024. 

But four lawsuits relying on Veracity Petitions’ software were voluntarily dismissed after opposing counsel on both sides of the aisle found the majority of allegedly invalid signatures were not invalid at all. 

In one case, 77 of the 103 claims of allegedly invalid signatures were found to be baseless. In another, a county report found of the 820 allegedly invalid signatures, 449 were valid. 

Andy Gaona, a former election attorney who defended a challenge against Marlene Galán-Woods in the 1st Congressional District in 2024, wrote in a letter to Spencer that the “entire case (along with others brought by your firm) was predicated on the abject failure of Veracity Petitions’ software platform.”

Spencer and Johnson did not respond to a request for comment. But as for this cycle, Veracity Petitions is still around. 

Per the website, the company continues to utilize signature recognition software to detect invalid signatures, with attorney and reviewer validation deployed before issuing a final report. 

“With Veracity Petitions’ software, attorneys are now able to take customers’ petitions and process thousands (or even hundreds of thousands) of signatures — accurately — in a fraction of the time,” the website reads. “Customers can expect the right turnaround time to make key campaign decisions and meet litigation deadlines.” 

How petition review software fares across the board this cycle remains to be seen, and paper candidate petition challenge could be more sparse going forward as candidates opt to use E-Qual, the secretary of state’s online petition signature platform that verifies a voter on the spot.

But in any case, attorneys stress the continued need for a human stopgap. 

“People who sign with terrible handwriting or have spelling mistakes are troublesome for things that care about details, like computers,” Langhofer said. “It’s just a really, really hard problem to solve. I don’t think it’s an imminent threat to human reviewers.”