The information architecture of websites is the most important remaining source of usability problems. Therefore, this research explores automated cognitive computational analysis of the information architecture of large websites as a basis for improvement. To support goal-specific analysis, an enhanced model of web navigation was implemented with a novel database-oriented approach. Web navigation was simulated on the information architecture of two large sites. With the improved labeling system of the information architecture, simulation results showed a significant reduction in navigation problems. The results of two experiments demonstrate that sites with improved information architecture result in better outcomes of user information retrieval. Our database-oriented approach is extensible, allowing non-goal-specific analysis, modeling of nontext media content, and analysis of the organization- and navigation systems of information architectures.