Systems and methods are provided for generating texts appropriate for a reading level of an individual. An existing exam unit is accessed, wherein the existing exam unit includes a reading passage and a plurality of questions related to the reading passage. The plurality of questions are filtered based on a criterion to form a subset of questions. A first difficulty score is determined based on the reading passage. A second difficulty score is determined based on the subset of questions. A correlation between the first difficulty score and the second difficulty score is determined, and a text is generated that is appropriate for a reading level of an individual based on performance of the individual on the exam unit.