<p>Triaxial braided (TB) composites are widely used in complex structures due to their high design flexibility and superior axial mechanical properties. However, their mechanical properties are highly sensitive to tow architecture, and the intricate interlaced configuration makes accurate prediction challenging. Most existing analytical models are based on the idealized TB-basic (TB-B) architecture, which assumes straight axial tows, thereby neglecting unintended tow interlacement during braiding. As a result, these models fail to capture axial tow undulation and its influence on mechanical properties, as well as the architectural transition from TB-B to tow-interlacement-induced TB-deformed (TB-D) composites. In this study, an analytical model was developed to predict the elastic modulus of TB composites by explicitly incorporating the paths of both braided and axial tows, as well as the pattern transition between TB-B and TB-D architectures. Cross-sectional geometry was examined using an optical microscope to identify the axial tow path and confirm the predicted undulation. Tensile tests were performed for validation. The proposed TB-D model predicted the axial modulus with an error of less than 10% and showed improvement over the TB-B model. Parametric studies revealed that axial tow undulation reduces axial modulus, particularly at higher braiding angles. In contrast, its influence on the transverse modulus was negligible. Furthermore, the analytical model was used to generate a large-scale dataset for AI training, and sensitivity analysis was conducted to identify the dominant geometric parameters governing the elastic moduli. These results provide a realistic framework for predicting the mechanical behavior of TB composites.</p>