Despite its evolutionary and ecological relevance, the mode of polyploid origin has been notoriously difficult to be reconstructed from molecular data. Here, we present a method to identify the putative parents of polyploids and thus to infer the mode of their origin (auto‐ vs. allopolyploidy) from Amplified Fragment Length Polymorphism (AFLP) data. To this end, we use Cohen's d of distances between in silico polyploids, generated within a priori defined scenarios of origin from a priori delimited putative parental entities (e.g. taxa, genetic lineages), and natural polyploids. Simulations show that the discriminatory power of the proposed method increases mainly with increasing divergence between the lower‐ploid putative ancestors and less so with increasing delay of polyploidization relative to the time of divergence. We apply the new method to the Senecio carniolicus aggregate, distributed in the European Alps and comprising two diploid, one tetraploid and one hexaploid species. In the eastern part of its distribution, the S. carniolicus aggregate was inferred to comprise an autopolyploid series, whereas for western populations of the tetraploid species, an allopolyploid origin involving the two diploid species was the most likely scenario. Although this suggests that the tetraploid species has two independent origins, other evidence (ribotype distribution, morphology) is consistent with the hypothesis of an autopolyploid origin with subsequent introgression by the second diploid species. Altogether, identifying the best among alternative scenarios using Cohen's d can be straightforward, but particular scenarios, such as allopolyploid origin vs. autopolyploid origin with subsequent introgression, remain difficult to be distinguished.