The paper addresses the variability of Czech phrasemes, i.e. semantically non-compositional multiword units, in current use represented by corpora, the variability being the result of linguistic creativity on the part of text authors. It also asks what, in fact, identifies a phraseme. A basic, original phraseme has a certain meaning that cannot be inferred from the meaning of its components, and if it is modified, made more topical and up-to-date, either the original meaning is entirely or partially preserved, or the modified phraseme acquires a totally new meaning. Some phrasemes allow for multiple modifications, while others are more rigid. The article examines different types of lexical/syntactic/morphological/semantic alteration of basic phrasemes. In addition to lexical variations, the focus is mainly on syntactic and morphological changes, and on the question as to whether the chosen syntactic means of expressing semantic shifts have an impact on the potential for a creative treatment of the phraseme. In order to identify the variants of a phraseme with the phraseme itself, we introduce the term phraseme nucleus and outline a partial solution to the phraseme variability problem — designing a lexical database of multiword units (including phrasemes) containing entries sufficiently flexible to at least partially capture the variability of phrasemes.