A study that examined whether state-level Supplemental Nutrition Assistance Program (SNAP) websites provide equitable access to information through chatbots found that most states lack chatbot access, and multilingual access is even rarer, creating significant equity gaps.
Out of the 56 SNAP websites the researchers reviewed, only 11 (19.6%) had a chatbot, and just 4 of those supported Spanish. Montana was the only state whose chatbot included an eligibility screener, a feature that could greatly improve user experience.
"We hypothesized that chatbot design, particularly language support and interpretive capabilities, may create invisible barriers for non-English speakers and marginalized users, limiting their ability to access essential public services," said lead author MJ Salas, a Ph.D. student at the Rutgers University School of Communication and Information.
The study, "Invisible Barriers: Chatbots, Language, and Access to SNAP Information," by Salas, Assistant Professor of Communication Yonaira Rivera, and Associate Professor of Library and Information Science Vivek Singh, was published in the Proceedings of the Association for Information Science and Technology in October 2025. The study was partially supported by a grant from the New Jersey State Policy Lab at Rutgers.
"We emphasize the need for public-facing technologies to be critically examined—not just for functionality, but for their role in shaping equitable access to information."
"These findings matter because they reveal how technology can unintentionally exclude vulnerable populations," Rivera said. "By improving chatbot design with crosslinqual parity, better visibility, and nuanced interpretation, public agencies can make essential services more accessible. For the public, this means easier and fairer access to SNAP benefits, reducing food insecurity and promoting health equity."
When people chat online with websites (for example, websites for government agencies, retail stores, banks, airlines, hospitals, etc.), they are most likely communicating with a chatbot. An AI chatbot is a computer program that uses AI technology to be able to understand and respond to people's questions. While AI chatbots are designed to sound human and trustworthy (their designers assign them "personalities") they can provide incorrect information if their databases lack information (or their databases include false or misleading information) or the chatbot misunderstands what the person is asking for or how they ask it.
SNAP is the largest federal nutrition assistance program, serving millions of low-income families. As agencies adopt digital tools like chatbots, these systems become gatekeepers to critical information. If chatbots lack multilingual support or fail to interpret user queries, they reinforce information poverty, a condition where marginalized communities cannot access useful or trustworthy information even when it is technically available. This deepens existing inequities in food security and public health.
To conduct the study, the researchers audited all 50 U.S. states and 6 territories’ SNAP websites using the USDA directory and supplemental searches. They manually reviewed each site for chatbot presence and features. Using an inductive, qualitative coding approach, they analyzed visibility, language support, interpretive nuance, and privacy messaging.
"This grounded approach enabled us to identify both expected and unexpected barriers to equitable access," the authors said.
Their findings also revealed additional problematic issues: chatbots often fail to interpret nuanced questions, especially in Spanish, and visibility and privacy messaging were inconsistent. Some chatbots were easy to find and offered privacy warnings, while others were hidden or lacked user feedback options.
"SNAP is the largest federal nutrition assistance program, serving millions of low-income families. As agencies adopt digital tools like chatbots, these systems become gatekeepers to critical information. If chatbots lack multilingual support or fail to interpret user queries, they reinforce information poverty."
"Our initiative seeks to address these disparities by developing policy and design guidelines for equitable chatbot deployment," Singh said. "Our focus on language equity—specifically the inclusion of Spanish alongside English—reflects the linguistic diversity of the U.S. population, where Spanish is the most widely spoken non-English language. By auditing existing SNAP chatbot implementations, we aim to identify both the best practices and critical gaps in current systems to inform more inclusive digital service design."
To address these gaps, the authors said chatbot systems should be developed with a focus on equity and inclusion. This means chatbots should provide balanced support for multiple languages during training and testing, offering clear and accessible interfaces, and integrating privacy notifications and real-time feedback mechanisms to foster user trust. Such design choices enhance usability while ensuring chatbots act as inclusive digital access points rather than perpetuating existing inequalities.
"We emphasize the need for public-facing technologies to be critically examined—not just for functionality, but for their role in shaping equitable access to information," the authors wrote.
Learn more about the Ph.D. Program, the Communication Department, and the Library and Information Science Department at the Rutgers School of Communication and Information on the website.
