Unemployment insurance (UI) take-up is relatively low in the United States. We implement a large-scale field experiment among 50,000 likely unemployed individuals to study the causes and labor supply implications of incomplete UI take-up. Informational letters increased applications and receipt, with effects concentrated among low-wage workers. Rejection rates among treated applicants also increased: this suggests that the letters primarily reduced learning costs rather than improved eligibility beliefs. Randomized messages aimed at reducing free-rider stigma induced more applications, primarily among high-wage job seekers. Although prior work finds that more generous UI slows job finding, our take-up intervention modestly increased reemployment, as work-search requirements hastened job finding among recipients but also screened out applicants who were unwilling or unable to verify their search. We develop and estimate a structural job search model calibrated to the reduced form-experimental results to quantify these frictions and show that lower search-compliance costs yield the largest welfare gains for unemployed workers.