Voting technology works in three stages: a device records the voter's choices, a tabulator counts the records, and a paper trail lets election officials check the count against something a human can read. The security of the whole chain rests on that last stage. Machines can fail or be misconfigured; a paper record, sampled and compared by hand, is what makes the result verifiable.
The qualification matters. Technology does not make an election trustworthy on its own. Certification, testing before use, and post-election checks do. A voter who wants to follow the process can start with public tools: the Tennessee Secretary of State's election day page points voters to sample ballots, polling locations, and results going back decades, and the state's GoVoteTN search lets residents look up their registration and ballot information directly.
How does a voting machine actually record a vote?
Two device types dominate. A ballot-marking device lets a voter select choices on a screen and prints a paper ballot showing those choices; the voter reviews the paper before it is cast. A scanner then reads the marked paper and stores the vote. In a second arrangement, a voter fills out a paper ballot by hand and feeds it to the scanner, which tabulates it on the spot.
Either way, the design principle is the same: the voter's intent ends up on paper first, and the machine's count is a convenience layered on top of that paper. Devices that only record votes electronically, with no voter-verifiable paper record, are the category election security experts have long urged jurisdictions to retire, because there is nothing independent to recount.
What happens to the ballots after the polls close?
Precinct scanners store their counts on removable media, which are transported under chain-of-custody procedures to a central counting location. There, a tabulation system aggregates the results from each precinct. Absentee and mail ballots follow a parallel path: they are received, verified against the voter rolls, opened, and scanned, with the envelopes retained.
Every step is supposed to be observable. Poll watchers and the public can generally observe the transport, the scanning, and the counting, under rules each state sets. The mechanics are public even when the software is not, and the published procedures, not the press conference, are where a skeptical reader should look.
Why does the paper trail matter so much?
A paper record separates what the voter intended from what the machine reported. If a scanner is miscalibrated, its software misbehaves, or its memory card fails, the paper ballots still exist and can be counted by hand. That is the entire argument for voter-verifiable paper audit trails, and it is a structural argument, not a partisan one.
It also bounds what technology can be blamed for. A machine cannot silently change an outcome if officials compare a hand count of sampled paper ballots against the machine's tally and the numbers agree. Where they disagree, the audit expands until the discrepancy is explained. The paper is the referee.
How does a machine earn its place in the polling place?
Before any device touches a real ballot, it goes through certification and testing. Certification checks that the system meets federal and state standards for accuracy, accessibility, and security. Then, before each election, jurisdictions run logic and accuracy tests: they feed the machines known ballots and confirm the machines count them correctly, in public.
We cover the certification path in detail in our companion piece, Voting System Certification: How a Machine Earns the Ballot Box. The short version: no vendor ships a system straight to a county. The machine has to pass review, then pass testing in the jurisdiction that will use it, and the test results are part of the public record. We covered a connected angle in Voting System Certification: How a Machine Earns the Ballot Box.
What can an audit prove — and what can it not?
A post-election audit samples a subset of ballots, counts them by hand, and compares the result to the machine count. If the sample matches within tolerance, the audit confirms the tabulation worked as intended for that contest. Risk-limiting audits, which several states have adopted, widen the sample automatically when the initial comparison looks off, so the check scales with the risk.
What an audit cannot do is retroactively verify an election conducted entirely without paper. It also does not examine voter eligibility or registration; those are separate controls on separate records. Our analysis section has argued before that the audit is the load-bearing wall of election integrity, and the claim holds here: it is the one step that converts a machine count into a humanly verified one.
What this means for voters
The practical takeaway is narrow and useful. Verify your registration and polling place before election day; the Vote.org polling place locator aggregates lookups by state, and state portals such as Tennessee's do the same directly. Review your sample ballot in advance, since down-ballot races are where long waits and voter confusion concentrate. And when you use a ballot-marking device, read the printed paper before casting it, because that review is the control the whole system depends on.
None of this requires trusting a vendor's marketing or a social media claim. The documents that matter are the certification records, the published testing procedures, and the audit results, all of which jurisdictions release. Voting technology is best understood not as a black box but as a chain of checks, each one public, each one bypassable only if nobody is watching.
Where the technology debate goes next
The open questions are about replacement cycles and access, not mystery. Machines age, standards tighten, and counties with small budgets face the same procurement bottlenecks that slow every part of government technology purchasing, a pattern familiar from our reporting on why government websites fail. The evidence so far supports a simple conclusion: jurisdictions with paper ballots, pre-election testing, and routine audits have a verifiable system, whatever device sits on the table. Jurisdictions without those checks are asking voters to take the machine's word for it. That is the distinction worth watching, and it is visible in public records in every state. Readers following this should also see Digital Services: Why Government Websites Fail and What Actually Fixes Them.
