Carbon fiber reinforced composites possess excellent mechanical properties such as high stiffness and strength in relation to their weight. However, their anisotropic nature and the multi-step production process consisting of textile manufacturing, preforming and matrix material injection lead to several defect sources. Non-destructive testing can decrease the amount of waste and simultaneously provide feedback to optimize process parameters. One of the technologies that are frequently employed is eddy current testing (ECT) because it can evaluate both textile preforms and infiltrated composites. However, current ECT cannot provide reliable information on the depth of defects and thus the affected reinforcement layer cannot be determined. Multi-frequency ECT in theory allows to generate depth-dependent signals but currently there is lack of evaluation methods to process these signals. Here we propose and validate a method based on an analytical approach to reach a depth-resolved image of the sample. To leverage the frequency-dependent penetration depth, an ECT device, which acquires signals at 4 different frequencies is used. Several carbon fiber reinforcement configurations with different defect sizes and orientations are setup and measured. The validation is carried out on carbon fiber textile stacks of a thickness of up to 1.2 mm. Results demonstrate successful layer separation for materials with high conductivity contrasts, though limitations exist in resolving fiber orientations within individual layers. The method shows promise for detecting and localizing defects with significant conductivity differences from the base material.