Background
New Jersey agencies increasingly use artificial intelligence (AI) tools to improve access to public information and services. During the pandemic, the State launched a COVID-19 Information Hub to centralize guidance and help residents find resources quickly. Since 2024, New Jersey has also provided a secured internal AI Assistant and training so staff can produce plain-language content, analyze resident feedback, and improve service workflows. In parallel, the State adopted a language access law that requires Executive Branch entities providing direct services to translate vital documents into multiple languages and to offer interpretation services, setting a high bar for multilingual digital services.
Definition of chatbots
A chatbot is a software application that uses natural language processing to understand a user’s question and generate a relevant answer. Modern public-sector chatbots can ground answers in authoritative documents and present plain-language guidance, but they require guardrails for accuracy, privacy, and multilingual equity.
Purpose
This report examines how chatbots are used to deliver public information, where language equity gaps appear, and how a bilingual, retrieval-grounded prototype can support more equitable access to New Jersey’s Supplemental Nutrition Assistance Program (SNAP) information for English- and Spanish-speaking households. It aims to inform policy and design choices that align with statewide AI guidance and language access requirements.
Methods overview
The report uses three complementary methods: (1) a multi-state audit of 56 state/territory SNAP websites to assess chatbot availability, features, language support, and performance; (2) semi-structured interviews with New Jersey stakeholders to identify desired functions, language needs, and expectations for human escalation; and (3) design and testing of a bilingual, retrieval-grounded chatbot prototype evaluated for accuracy, readability, and parity across English and Spanish.
Key findings
New Jersey’s SNAP website currently lacks a resident-facing chatbot, creating an opportunity for a bilingual assistant that aligns with statewide equity requirements. Some of the other states’ chatbots are often visible and easy to access but struggle with nuanced eligibility questions and perform inconsistently in Spanish, underscoring the need for retrieval grounding and dedicated bilingual quality assurance. New Jersey’s existing AI guidance, training infrastructure, and language access law position the State well for a measured and accountable pilot.
Implications
A retrieval-grounded, bilingual chatbot can help New Jersey meet its multilingual service obligations, reduce confusion, and improve trust while providing clear pathways to human assistance. Continuous evaluation, feedback integration, and transparent communication will be essential for sustainable and equitable deployment.