This paper uses nationally representative survey data of adults Internet use in Britain to examine current patterns in the uptake of lifelong learning via the Internet. We develop and test a model that accounts for structure, agency and outcomes using structural equation modelling to address two questions: (1) how structure (as measured by age, gender, SES, Education and ACORN) is related to personal and capital enhancing outcomes of learning online; and (2) how agency (as measured by digital skills and engagement with online learning) mediates this relationship. We demonstrate that social structure remains an important factor in understanding patterns of uptake and outcomes of online learning, alongside an individual's agentic behaviours. We suggest that countries such as the UK, which have become overly focused on individual interventions to increase the uptake of lifelong learning via the Internet, are going in the wrong direction. Such interventions have failed in the past, and we suggest that they will continue to do so unless policy makers reconceptualise lifelong learning and the Internet in ways that take social structures into account.