The paper addresses the variability of Czech phrasemes. This variability is verified in the corpus and is captured in the upcoming multi-word expression database, which should enable to identify and search for these expressions, including idioms and their variants in the corpus.