Leveraging AI to prevent international child abductions

About State

The U.S. Department of State (State) leads America’s foreign policy through diplomacy, advocacy, and assistance by advancing the interests of the American people, their safety and economic prosperity. The Department’s Bureau of Consular Affairs (CA) serves more than 30 million customers each year in the United States and every country in the world. As part of the CA’s Overseas Citizens Services (OCS) Transformation Team, a U.S. Digital Corps Data Science and Analytics Fellow works to improve the Children’s Passport Issuance Alert Program (CPIAP), a service that OCS provides to parents & guardians to prevent international child abductions.

The challenge

Initial research on OCS’s Child Abduction Prevention Program revealed a critical problem: despite being the primary source of information about the program, its webpages were difficult to navigate, confusing, and repetitive. As the team planned a comprehensive redesign, it needed baseline metrics to measure how improvements would enhance user experience. However, a Google Analytics configuration error prevented any visibility into how parents and guardians were actually interacting with these resources.

Compounding the issue, the Prevention Team spent significant time manually combing through thousands of emails from concerned parents and guardians, repeatedly addressing the same questions. This approach pulled frontline staff away from more complex, time-sensitive cases, further delaying information to parents and guardians.

The approach

The Fellow began by diagnosing and correcting the Google Analytics configuration error that had been preventing any visibility into user behavior on CPIAP webpages, establishing the reliable baseline metrics necessary for measuring the impact of the planned improvements.

Recognizing that the volume of similar questions received via email indicated deeper usability issues, the Fellow saw an opportunity to address the root cause. She hypothesized that many repetitive questions could be eliminated by improving the web content itself, making critical information easier to find and understand so parents and guardians would not need to reach out for help.

To validate this hypothesis and understand the full scope of user pain points, the Fellow independently developed and deployed an advanced large language model (LLM) between June and July 2025. The AI system analyzed over 100,000 customer service emails, efficiently categorizing inquiries and uncovering patterns that would have required months of manual review. This analysis revealed which aspects of the enrollment process created the most confusion and friction for families, providing the Prevention Team with actionable, data-driven insights to guide the website redesign and content improvements.

The impact

The Fellow’s work transformed how OCS understands and serves families through CPIAP. With functioning web analytics now in place, the Prevention Team can measure how users navigate each of the different webpages, identify where they might be dropping off, and continuously refine the overall digital experience. The insights from the email analysis are directly informing website redesign efforts, enabling the team to proactively address common questions through improved content rather than reactive email responses.

Parents and guardians benefit from faster access to clear, comprehensive information about protecting their children from international abduction. By reducing confusion and streamlining the enrollment process, this work expands the program’s reach and helps more families take preventive action. Meanwhile, frontline staff can dedicate their time and expertise where it matters most: handling complex cases that require immediate intervention to keep children safe.