As the voting advice application (VAA) market has grown increasingly competitive, developers have focused on innovation, including specialised VAAs that focus on specific topics. This article evaluates the analytical potential of a “single-party VAA” called BSW-O-Mat, a website developed to assess alignment with the newly founded German party Bündnis Sahra Wagenknecht (BSW), attracting over 50,000 participants. Single-party VAAs face compounded selection bias; beyond the typical overrepresentation of educated politically engaged users, they disproportionately attract voters interested in the focal party. Despite over 50,000 responses, questions remain about data validity for estimating voter preferences. To address the usefulness of VAA data for analysing the new party’s supporters, this study examines convergence validity by comparing the average marginal effects from logistic regression models based on VAA data with those from the 2025 German Longitudinal Election Study’s (GLES) probability sample. Despite substantial overrepresentation of potential BSW supporters in the VAA sample, convergence in the direction of effects was observed across datasets. Effect correlations reached r = 0.65 (increasing to r = 0.87 when excluding non-converging effects), though the VAA dataset tends to overestimate effect sizes. Notably, even analysing substantially underrepresented AfD voters (5.5% in VAA vs 19.8% in GLES) yielded results largely consistent with probability-sample models. These findings demonstrate that the VAA data can provide valid insights when appropriately weighted and interpreted in light of self-selection biases.