A handful of studies have claimed that error detection is improved by a proofreader’s prior encounter with the text to be scanned for errors. In these studies, however, the beneficial effect of text familiarity on proofreading has been obtained via surface encoding tasks (prior reading or proofreading). This raises the question of whether the effect is dependent on the type of encoding operations performed on the text prior to proofreading. In Experiment 1, familiarization required that subjects read an error-free passage and then either type the passage verbatim (surface encoding) or write an essay by relying on the information contained in the passage (deep encoding with generation). In Experiment 2, subjects generated sentences (deep encoding with generation), assessed whether sentences described drawings (deep encoding), or typed sentences (surface encoding). Familiarity, irrespective of the nature of the encoding operations, improved proofreading times but only surface encoding or deep encoding without generation made proofreading more accurate. In Experiment 3, text generation fostered greater subjective familiarity than either surface or deep encoding alone. It is argued that this form of enhanced familiarity may lead to expectancy effects in proofreading performance.