This study examines how far the lexical bundles that occur in graded readers are influenced by simplified language, comparing them quantitatively and qualitatively with those occurring in a corpus of authentic prose fiction. Phrasal language found in the graded readers is also evaluated using Martinez and Schmittâs (2012) PHRASE list. The results are largely encouraging, showing that lexical bundles occur with greater density in graded readers than authentic fiction, that they largely reflect authentic language use, and that most of the phrases deemed to be important and useful are represented. However, differences between B1 and B2 level readers indicate that a higher degree of simplification affects both the range and grammatical type of lexical bundles. Non-transparent lexical bundles, despite being composed of very frequent words, were under-represented in the readers, particularly at B1 level. It is concluded that while graded readers are a valuable source of exposure to lexical bundles, the under-representation of frequently-used opaque phrases points to the need for a more systematic approach to their inclusion.