The ALEKS assessment uses open-ended problems (no multiple-choice questions). It is an adaptive assessment; that is, problem types are selected based on all the previous answers the student has given. It is impossible to predict which types of problems will appear, or in what order. Moreover, the problems themselves are generated algorithmically, with randomly-selected values (as is also the case in the Learning Mode). Thus, one cannot "learn the assessment" or "teach to the assessment," and cheating is impossible. In the unlikely event that two students sitting next to one another were given the same problem-type at the same time, the problem parameters and values would almost certainly be different, and so would the correct answer. Despite this, certain assessments must be supervised, such as the initial, midterm, and final assessments in a course. Without supervision, students could use a textbook, receive systematic help, or have someone else take the assessment in their place. This point is critical where assessment results are used for purposes other than those internal to the system. (There is no reason for a student who has begun using ALEKS to cheat on a "progress" assessment, as this will simply cause the system to suggest problems that are too difficult, and thus hinder the student's own work.)

As noted, the student takes an initial assessment immediately following completion of the Tutorial (See Sec. 3.9). When an assessment begins, the student is clearly informed it has begun. Next a series of problems is posed to the student. The student provides the solution to each problem using the Answer Editor (or clicks "I don't know"). In the Assessment Mode, the system does not inform the student whether the answer just given was correct or not. The assessment continues until the system has determined the student's precise knowledge of the domain, at which time the assessment ends and a report is presented to the student. The number of questions asked cannot be known in advance, although consistency of effort and attention seem to contribute to shorter assessments.