Selection signatures are genomic regions that carry the imprint of past selection events, offering key insights into the evolutionary forces and human interventions that have shaped the genomes of domesticated species. Over the past two decades, numerous methods have been developed to detect such signatures, including site frequency spectrum (SFS)-based statistics, haplotype and linkage disequilibrium (LD)-based approaches, population differentiation metrics, and, more recently, machine learning and deep learning techniques. Consequently, selection signature analyses have become powerful tools in livestock and crop genetics for uncovering genes associated with economically important traits such as growth, reproduction, disease resistance, and adaptation to environmental stressors. However, most studies to date have focused on detecting selection signatures between populations, while the detection of genomic regions under selection within populations remains relatively unexplored. To address this limitation, we developed TRACE (Tracking Recombination and Admixture Captures Evolution), a novel approach that identifies intra-specific selection signatures by incorporating individual admixture proportions in genome-wide association studies (GWAS). The performance of TRACE was first evaluated using simulated data (1,000 individuals and 100,000 SNPs), where the method successfully identified a quantitative trait locus (QTL) whose effect is strongly associated with individual admixture proportions, demonstrating its ability to detect genomic regions underlying admixture-driven selection. We then applied TRACE to the Franches-Montagnes horse breed, which has experienced multiple historical admixture events during its formation, including more recent contributions from Warmblood, purebred Arabian, and Shagya Arabian horses. Individual admixture proportions were estimated using high-density SNP genotype data. The dataset included 522 Franches-Montagnes horses (including 44 old-type individuals), 514 European Warmblood horses from Sweden and Switzerland (including a stallion used for crossbreeding in 1990), 136 Arabians, 32 Shagya Arabians, and 64 Thoroughbreds, as pedigrees of introgressed Warmblood stallions indicated Thoroughbred ancestry. By using the Warmblood admixture proportion of Franches-Montagnes horses as the GWAS phenotype, we identified two selection signatures surpassing the genome-wide Bonferroni-corrected significance threshold (p_BONF HPGD and PAK5 genes, respectively. PAK5 has previously been detected in selection signature analyses of Swedish Warmblood horses and is implicated in neuronal outgrowth, which is essential for locomotion and cognitive function. Mutations in HPGD are associated with digital clubbing, a key feature of primary hypertrophic osteoarthropathy (PHO). In horses, the selection signature identified on ECA2 lies within a previously reported QTL associated with radiological alterations in the contour of the navicular bone in Hanoverian Warmblood horses. Overall, our results demonstrate that TRACE enables the detection of selection signatures within populations by leveraging admixture proportions as quantitative phenotypes, thereby capturing recent or ongoing selection processes driven by modern breeding programs or local adaptation.