This paper examines the digital divide in internet use in general and whether there may be digital divides in internet use for specific purposes (leisure, improving human capital and obtaining goods and services). It uses a unique dataset which covers the entire clickstream of almost 20,000 internet users in the five largest EU economies during 2011. Our main finding is that, conditional on internet adoption, the digital divide in internet use has been reversed. Low-income internet users spend more time on the internet than high-income users. In addition, we find the effect of income on internet use is not affected by employment status of the internet users and we discuss several possible explanations for this result. There is some evidence of an education-based digital divide in the ability to use human capital and goods & services websites. Tertiary education has a negative effect on time spent on leisure websites and a positive effect on time spent on human capital and goods & services websites. Using quantile regressions, we find that the negative effect of income on time spent on internet and the positive education effect for human capital and goods & services websites holds for the entire conditional distribution of these online activities. Moreover, these effects are stronger for more intensive internet users.