Textbook creation has become big business for an increasingly small number of publishers in the world of ELT. Whilst dictionaries and reference works have become very heavily influenced by corpora in the last thirty years, the same phenomenon has not been observed in language textbook creation, which has largely remained the domain of writers’ intuition. After briefly looking at the methodological questions behind the use of either authentic corpus-based or intuitively created teaching materials, and why corpus data uptake to textbooks has been slow, this article asks the question ‘What can corpora tell us about textbooks?’ A broad overview is given of the comparative studies that have used corpora to examine the accuracy of English language teaching textbooks. The unique challenges related to creating textbook corpora are then noted, followed by a case study that examines the differences between actual business English and a corpus of business English textbooks. Finally, recommendations are made for future research, and how both teachers and publishers alike can enable corpus data to play a greater role in the future development of this important area.