This paper describes the development of web-based software at a university in Spain to help students of EFL self-correct their free-form writing. The software makes use of an eighty-million-word corpus of English known to be correct as a normative corpus for error correction purposes. It was discovered that bigrams (two-word combinations of words) present in a student's writing but which are not found, or which are suspiciously rare, in the normative corpus are likely to contain errors. The program highlights such bigrams in compositions, and guidance is provided to help students decide if they have indeed made a mistake, and if so, how to correct it. A cohort of students who volunteered to trial the software for a month reported positively on their experience; it had helped them find mistakes in their own and their peers' writing and had greatly accelerated the self-correction process.