Holding elections has become a global norm, even in autocracies; at the same time, there is mounting evidence to suggest that flawed or failed elections pose serious risks for political stability, legitimacy, and participation. Scholars and practitioners alike increasingly see domestic election monitoring groups to be a partial remedy to electoral malpractice. At least half of elections globally are monitored by such groups and large sums of international aid spent on them. A small but growing empirical research literature examines the effects of domestic observers on electoral integrity. These studies rest on important but unsubstantiated assumptions about the goals of election watchdogs. For instance, the account of ‘self-enforcing democracy’ expects that observers aim to deter or displace election fraud (Ichino & Schündeln 2012; Hyde & Marinov 2014). Other authors posit that election watch groups want to build associational social capital and socialize political elites into norms of electoral integrity (Lean 2013). To date, however, little systematic and comparative research has empirically investigated these assumptions. This study addresses this gap and asks which theory of change underlies the increasingly widespread activities of citizen election monitors. What do citizen observer groups themselves see as their strategic goals? The study draws on an original dataset of 1,000 citizen election wat NGOs in 114 countries, an organizational survey of 200 of these groups, and in-person interviews with election watch activists from 20 countries, to address this open question. It argues that domestic election monitors primarily play a role as agenda builders, shaping the public debate about electoral integrity, aiming to affect procedural, legal, and behavioral changes in the long run. The study thus draws into question some of the established assumptions of the literature on election monitors and raises the question whether the common empirical tests are suitable to capture the intended effects of domestic observers.