International audience A major preoccupation of English for specific purposes, however defined, is to provide fine-grained analyses of targeted varieties of English. The use of appropriately specialised corpora can clearly help with such descriptions; the question then is what to do with the results. Much work has been carried out by experts upstream, for dictionaries and other reference tools, for coursebooks and other materials, with particular interest in frequency lists of words and phrases for ESP and EAP. The corpus input, however, is often virtually invisible to the end-user.This paper examines what can happen when corpus tools and techniques are put into the hands of teachers, learners, translators and professionals in what is commonly known as data-driven learning. While this has been called ‘radical’ and ‘revolutionary’, an alternative is to emphasise how it builds on common pedagogical underpinnings (e.g. an inductive approach to authentic documents) and existing practices (e.g. web searches), thus enabling better integration as an everyday activity for different types of users. Moving on to corpora, three sets of examples are given, all with free, reliable, user-friendly tools. Beginning at the level of text, comparison to a reference corpus can help pinpoint its specificities, thus encouraging noticing, and help the user decide what to focus on. Second, general-purpose corpora can provide considerable help for even quite specific points that learners have in their ESP work, especially where the corpora can be queried by subsection. Third, it is increasingly quick and straightforward to build one’s own corpus, closely oriented towards the user’s individual needs in relation to a specific variety. We finish with a survey of research in the field based on a comprehensive meta-analysis, concluding that corpora can prove extremely useful in a variety of ways, especially perhaps as a reference resource for ESP writing and revision.