The ability to use the internet frequently is likely to provide a useful means of engaging with society and using services in later life, yet older people are at the most likely to suffer digital exclusion, with those of the oldest ages at the greatest risk. Using six waves (2002-2012) of the English Longitudinal Study of Ageing, we model cohort-specific patterns of frequent internet use for people aged 50 and over. Multilevel growth models are used to observe trajectories of internet use over the ten year period. Firstly, analyses are stratified by gender and wealth, and secondly we additionally test for health effects. The study finds cohort-specific differences in patterns of internet use. Rates of internet use increase faster among younger cohorts yet, despite initially increasing, begin to decline among older cohorts. Poor health is shown to be a key factor in shaping the trajectory of internet use over time. Rates of internet use are consistently lower for women than men and for those in poorer financial circumstances, independently of age cohort. The findings demonstrate the importance of ensuring older people can remain digitally included throughout later life, including after the onset of poorer health, especially as some of these individuals might benefit the most from some of the services the internet can provide.