Legibility of Summary-Style Printed Ballots
The ability of voters to review and verify their selections before casting their ballot is an important step in the voting process. This report explores the legibility and readability of summary ballots printed by ballot marking devices (BMDs) and the ability of optical character recognition (OCR) applications commonly found on mobile phones to voice (read) summary ballots.
The report looks at the visual presentation of the ballot to identify typographic elements that might make it easier to read visually, the feasibility of using OCR to allow blind or low vision voters to hear their ballot read accurately, and whether there is a relationship between the design elements that support both visual and OCR-assisted reading.
This research was published by the National Institute of Standards and Technology (NIST) as VTS 100-4, also available on NIST’s website.
Key findings
Despite the challenges of reading current ballots with OCR tools, many of the design problems have straightforward solutions when accessibility is prioritized. The analysis points to five design elements that matter most:
- Text size should meet the Voluntary Voting System Guidelines (VVSG) 2.0 requirements for printed ballots.
- Line spacing needs to be sufficient so that character ascenders and descenders don’t overlap and crowd the text.
- Clear layout is essential — header information should be visually separated from contest data, making it easy for voters to find where their selections begin.
- Contest separation using typography or design elements (white space, horizontal lines, bold text) helps voters and OCR readers process each race as a distinct unit.
- Candidate identifiers — such as party affiliation or election codes — should appear after the candidate’s name, not before it.
Beyond ballot design itself, the research points to alternative verification approaches worth exploring: the ability to have a marked ballot read back at the voting station, a review display at the ballot scanner before casting, and QR codes or similar encodings that could be more reliably read by personal assistive technology.
Making these improvements will require both lab testing with OCR tools and usability testing with blind and low-vision voters using their own devices in realistic polling place conditions.
About the research
This research was conducted by Lynn Baumeister, Whitney Quesenbery and Sharon Laskowski (NIST). It was published in January 2025 as part of the NIST Voting Technology Series (NIST VTS 100-4).
The research analyzed seven summary-style ballots collected from demonstrations by voting system vendors and election departments. This was an expert analysis of the ballots’ visual design and the effectiveness of OCR tools in reading them, not usability testing with voters
In the first part of the project, we conducted a visual legibility analysis, which examined measurable typographic characteristics of each ballot — including font size, line spacing, use of visual separators, reading patterns, and the placement of auxiliary information like party identifiers and election codes.
Next we conducted an OCR analysis, which tested how accurately five tools could read each ballot: a flatbed desktop scanner running the general purpose app, FineReader (used to establish a baseline), and four smartphone apps — KNFB Reader, SeeingAI, Voice Dream Reader, and Google Lens — held by hand over the printed ballots to simulate how a blind or low-vision voter might use personal assistive technology at a polling place.
Related resources
Visit our page on voting systems to find more resources about the usability and accessibility of voting systems.
© 2025 Center for Civic Design