In an
optical scan voting system, each voter's choices are marked on one or more pieces of paper, which then go through a scanner. The scanner creates an electronic image of each ballot, interprets it, creates a tally for each candidate, and usually stores the image for later review. The voter may mark the paper directly, usually in a specific location for each candidate, then mail it or put it in a ballot box. Or the voter may select choices on an electronic screen, which then prints the chosen names, usually with a bar code or QR code summarizing all choices, on a sheet of paper to put in the scanner. This screen and printer is called an electronic ballot marker (EBM) or
ballot marking device (BMD), and voters with disabilities can communicate with it by headphones, large buttons, sip and puff, or paddles, if they cannot interact with the screen or paper directly. Typically the ballot marking device does not store or tally votes. The paper it prints is the official ballot, put into a scanning system which counts the barcodes, or the printed names can be hand-counted, as a check on the machines. Most voters do not look at the machine-printed paper to ensure it reflects their choices. When there is a mistake, an experiment found that 81% of registered voters do not report errors to poll workers. No state requires central reporting of errors reported by voters, so the occasional report cannot lead to software correction. Hand-marked paper ballots more clearly have been reviewed by voters, but some places allow correction fluid and tape so ballots can be changed later. Two companies, Hart and Clear Ballot, have scanners which count the printed names, which voters had a chance to check, rather than bar codes and QR codes, which voters are unable to check. When scanners use the bar code or QR code, the candidates are represented in the bar code or QR code as numbers, and the scanner counts those codes, not the names. If a bug or hack makes the numbering system in the ballot marking device different from the numbering system in the scanner, votes will be tallied for the wrong candidates. Some places put a privacy screen over the BMD's display, so voters must be "aggressive" when pressing the touchscreen to register a vote. • In a 2023 election in Northampton County, PA, misprogramming switched text on paper between two judge retention contests, so a vote to remove or retain the first was labeled with the name of the second, and vice versa. The county briefly changed to hand-marked paper ballots, but ran out, and went back to the mislabeled BMDs. The county and court changed instructions throughout the day. • In a 2022 election in Marietta, GA, up to 157 voters were presented lists of contests which were partly in error, due to recent changes in district boundaries. • In a 2019 election in Philadelphia, PA, 40% of polling locations had problems with BMDs, including "touchscreens that were hypersensitive or that froze; paper voting receipts getting jammed in the machines; and panels opening on some machines to expose the equipment's electronic controls." • In a 2019 election in Northampton County, PA, BMD screens were hypersensitive or insensitive, from being "configured improperly in the factory" so voters had to keep trying to select their candidates. An instructional message was treated as a Republican candidate, so straight party votes omitted the real candidate.
Errors in optical scans Some scanners have a row of photo-sensors which the paper ballot passes by, and they record light and dark pixels from the ballot. Other scanners work by scanning the ballot to create an electronic image (i.e. a picture) of the ballot. Then the resulting image (picture) is analyzed by internal software to discern what votes were cast. In these types of vote tabulators, any defect in the scanning apparatus, such as dirt, scratches on the lens, or chads of paper, may mar the image of the ballot. These types of defects may produce long black lines, or white lines in the ballot image, which then may be misread by the software as an undervote (no votes for anyone), or as an overvote (a vote for every candidate). Some offices blow compressed air over the scanners after every 200 ballots to remove dust. Software can miscount or fail to count. If it fails or miscounts drastically enough, people notice and check. Errors can happen in the scanner or at the ballot printer. Staff rarely identify the person who caused an error, and assume it was accidental, not a hack.
2025 • In the 2025 general election, Rensselaer County, NY, formatted ballots and computers differently, and corrected with a hand count, so a state forest proposal gained 382 votes, a library levy gained 480 votes, and a rescue squad proposal gained 1,733 votes, changing outcomes for the latter two. The library had collected signatures from more voters for the levy than the computer results showed, and the county found a misprint. Citizen groups asked the state Attorney General to investigate. • In the 2025 general election, Cambridge, MA included 2,158 test ballots in its count. When they recalculated without them, one school board result changed.
2024 • In the 2024 general election in Sandoval County, NM, the memory device from one precinct became unreadable when a worker mishandled it, totals were inconsistent, and two voting machines were used without seals. Officials certified results anyway, and one candidate filed a complaint. • In the 2024 general election in Ashtabula County, OH, scanners mis-counted votes for levies, because the state Supreme Court changed the wording of the first state-wide ballot measure, so other measures shifted from where the scanners expected them on the page. The problem was discovered only when the county hand counted one measure because it was very close. Then the county hand counted other measures. Three measures changed from losing to winning. Other Ohio counties reported some ballots as unreadable, but did not hand count any ballots to check for shifted votes. • In the 2024 general election in Torrington, CT, broken tabulators and coding which did not match the layout of the printed ballots caused errors in counts of absentee and early ballots and ballots of voters who registered the same day they voted. Staff recounted these 7,000 ballots by hand. • In the 2024 general election in Centre County, PA, results from 13,000 scanned ballots did not upload to the central election computers, and were rescanned. • In the 2024 general election in Cambria County, PA, software and printing errors prevented counting votes on election day. Polls stayed open late and ballots were counted later. • In the 2024 general election in Nevada and Shasta Counties, CA, bar codes on the edges of ballots which tell machines the "style" or contests on the ballot were imprecisely printed ("ink overspray") by Runbeck, the ballot printing company, so scanners could not read them and process the ballots. They were hand-copied onto better ballots. • In the 2024 primary in Shasta County, CA, 100 ballots appear in the scanners' audit logs and not in final results, in a district with a contest decided by 14 votes. The logs show 20 manual time changes in 4 machines, half of which were logged exactly 60 seconds after a previous update. Seven were outside working hours, between nine at night and 4 in the morning, when video surveillance showed no one present. The county asked the US Justice Department to investigate. • In a 2024 primary in Utah, "adjudicated ballots were not reflected in the total results" in at least Tooele and Washington Counties, when more than a few adjudicated ballots were entered at once. • In a 2024 primary throughout Puerto Rico, parties found over 1,000 errors. The elections commission said the voting machines "incorrectly calculate vote totals". The election commission then "conducted a full vote tally and audited paper receipts from hundreds of ballot-counting machines." Dominion said the errors were in their software which exports counts from the voting system for public release.
2022 • In a 2022 election in Alameda County, CA, software was wrongly set to skip 250 ballots in round 1 of Ranked Choice Voting. These ballots had no first choice and should have been counted by their top choice. FairVote saw the error, and the county corrected result, which changed the winner in
Oakland Unified School District 4. • In a 2022 election in Monmouth County, NJ, software did not notice when staff uploaded six flashdrives of ballot images twice, which added more votes to candidates and switched a school board winner. It was caught by a council member who thought the numbers were too high, and made a public records request for detailed numbers, which showed more ballots than voters checked in. The part of the software designed to notice duplicates had not been installed, and there was no automated checking that the installation was right. • In a 2022 election in DeKalb County, GA, a candidate who actually won appeared to lose, after votes were not counted for her in some precincts, because another candidate withdrew, and programming did not categorize votes correctly. • in a 2022 election in Clackamas County, OR, scanners could not read more than 100,000 ballots on which the ballot printing company had printed a blurred bar code. The ballots needed to be hand-copied by teams of two onto correctly printed ballots. The problem was discovered May 3, for the May 17 election, and was not corrected until after the election. • in a 2022 election in Lancaster County, PA, scanners could not read 22,000 ballots on which the ballot printing company had put the wrong identification code. The ballots were hand-copied by teams of three onto correctly printed ballots. The state does not let mailed ballots be scanned before election day, so the problem was only discovered on election day.
2021 • in a 2021 election in Williamson, TN, precinct scanners had two errors: They misread QR codes, mistakenly classifying some ballots as provisional and not to be tallied, then kept that not-to-be tallied classification for later ballots, so large numbers of ballots were not included in precinct reports at some precincts. They were re-scanned correctly on a central scanner and checked by a hand-tally. The county rented another brand for the 2022 elections. • In a 2021 primary in New York City, 135,000 test ballots were not removed from the database and were included in preliminary counts for the mayoral primary. They were discovered because totals were higher than the number of voters, and corrected by removing them from the count. • in a 2021 election in Lancaster County, PA, scanners could not read 12,300 ballots on which the ballot printing company had put the wrong identification code. • in a 2020 election in Antrim County, MI, last minute updates to some ballots were not applied to all scanners, so the scanners had inconsistent numeric codes for different candidates and styles of ballots, causing errors of thousands of votes. Corrections happened in stages, leading to less and less confidence in the results. Results were eventually confirmed by a hand count. • In a 2020 election in Windham, New Hampshire, fold lines in the wrong places and dust on scanner sensors caused many fold lines to count as votes. • In a 2020 election in Baltimore, Maryland, the private company which printed ballots shifted the location of some candidates on some ballots up one line, so the scanner looked in the wrong places on the paper and reported the wrong numbers. It was caught because a popular incumbent got implausibly few votes, and corrected by hand-copying mailed ballots onto well-formatted ballots.
2010-2019 • In a 2019 election in Northampton county, Pennsylvania, the software under-counted one candidate by 99%, reporting 164 votes, compared to 26,142 found in a subsequent hand-count, which changed the candidate's loss to a win. • In 2018 scanners in various states had problems from humidity, rejecting too many ballots, rejecting staff passwords, delivery to the wrong locations, broken machines, power outages, screen calibration shifting votes to other candidates, and broken cartridges to start machines. • In a 2018 New York City election when the air was humid, ballots jammed in the scanner, or multiple ballots went through a scanner at once, hiding all but one. • In a 2016 Maryland election, a comparison of two scanning systems on the same ballots revealed that (a) 1,972 ballot images were incorrectly left out of one system, (b) one system incorrectly ignored many votes for
write-in candidates, (c) shadows from paper folds were sometimes interpreted as names written in on the ballot, (d) the scanner sometimes pulled two ballots at once, scanning only the top one, (e) the ballot printers sometimes left off certain candidates, (f) voters often put a check or X instead of filling in an oval, which software has to adapt to, and (g) a scratch or dirt on a scanner sensor put a black line on many ballot images, causing the appearance of voting for more than the allowed number of candidates, so those votes were incorrectly ignored. • In 2016 Wisconsin elections statewide, some voting machines did not detect some of the inks used by voters. • In a 2016 Rhode Island election, machines were misprogrammed with only one ballot style, though there were two. Results were surprising enough so officials investigated and found the error. • In a 2014 Stoughton, Wisconsin, election, all voters' choices on a referendum were ignored, because the scanner was programmed to look in the wrong spot on the ballot. It was corrected by a hand count. • In a 2010 New York election, 20,000 votes for governor and 30,000-40,000 votes for other offices were ignored, because the scanners overheated and disqualified the ballots by reading multiple votes in races where voters had properly only voted once.
Before 2010 • Errors from 2002 to 2008 were listed and analyzed by the
Brennan Center for Justice in 2010. • In a 2004 Yakima, Washington, election 24 voters' choices on 4 races were ignored by a faulty scanner which created a white streak down the ballot. • In a 2004 Medford, Wisconsin, election, all 600 voters who voted a straight party ticket had all their votes ignored, because the manufacturer forgot to program the machines for a partisan election. Election officials did not notice any problem. The consultant who found the lost 600 voters also reported a Michigan precinct with zero votes, since staff put ballots in the scanner upside down. • In a 2000 Bernalillo County (Albuquerque area), New Mexico, election, a programming error meant that straight-party votes on paper ballots were not counted for the individual candidates. The number of ballots was thus much larger than the number of votes in each contest. The software was fixed, and the ballots were re-scanned to get correct counts. • In the 2000 Florida presidential race the most common optical scanning error was to treat as an overvote a ballot where the voter marked a candidate and wrote in the same candidate. • Researchers find security flaws in all election computers, which let voters, staff members or outsiders disrupt or change results, often without detection. • Security reviews and audits are discussed
below.
Recreated ballots Recreated ballots are paper or electronic ballots created by election staff when originals cannot be counted for some reason. Reasons include tears, water damage, folds which prevent feeding through scanners and voters selecting candidates by circling them or other abnormal marks. Reasons also include citizens abroad who use the
Federal Write-In Absentee Ballot because of not receiving their regular ballot in time. As many as 8% of ballots in an election may be recreated. The estimate is $29 per voter ($203 million total) if all voters use ballot marking devices, including $0.10 per ballot for paper. The capital cost of machines in 2019 in Pennsylvania is $11 per voter if most voters mark their own paper ballots and a marking device is available at each polling place for voters with disabilities, compared to $23 per voter if all voters use ballot marking devices. This cost does not include printing ballots. New York has an undated comparison of capital costs and a system where all voters use ballot marking devices costing over twice as much as a system where most do not. The authors say extra machine maintenance would exacerbate that difference, and printing cost would be comparable in both approaches. Their assumption of equal printing costs differs from the Georgia estimates of $0.40 or $0.50 to print a ballot in advance, and $0.10 to print it in a ballot marking device. ==Direct-recording electronic counting==