Writers on library automation such as Borko and Lancaster foresee an end to human‐based indexing and classification. They anticipate a time when users will be able to direct their subject enquiries at machine‐held files of keywords extracted automatically from the ‘natural uncontrolled language of the document’. Borko considers that this will allow the user ‘to identify a few relevent items from among many thousands and display them on the video screen in seconds’. This paper reviews these claims, and then examines a genuine case where the computer followed the procedures proposed by these teachers. It also considers how a measure of vocabulary control, based on recommendations in some recent national and international standards, would have avoided the entirely irrelevant output produced by the machine.