Nowadays in France, microsatellites markers (MS) are used for parentage verification among all horse breeds. Studies have shown that parentage verification based on Single Nucleotide Polymorphisms (SNPs) are less prone to error, cheaper and have shorter processing times. Moreover, ISAG is working on a SNP panel to allow the transition from MS to SNP for parentage verification. The principal constraint of this transition is the cost of re-genotyping all sires and mares that are already MS-genotyped. In this study, we wanted to test a method to impute MS alleles from SNP haplotypes. A total of 6,295 horses from 3 breeds (Arabs, Selle-Français and French Totters) were used. They were MS genotyped for parentage verification and have been SNP genotyped with 54K to 670K chips. Haplotypes for markers within 500 Kb on either side of each MS were obtained using FImpute 3 by considering MS alleles as pseudo-snp. To determine SNP haplotypes that best fit a MS allele, we started by selecting 10 SNP on each side of the MS genotype, with the possibility of extending it if haplotypes matched with several MS alleles. The objective was to obtain a haplotype that always fitted one MS allele (or as little MS as possible). The haplotypes were built within breeds and across breeds. First analysis of the 9 alleles of MS AHT4 and the 166 haplotypes with 20 flanking SNPs showed mixed results. Problematic results from phasing, i.e. missing MS allele or more than one MS allele on the same haplotype, represented less than 0.5% of phased samples. Rare association (1 case) between MS allele and SNP haplotypes represented 1% of the phased samples. After removing these cases, SNP haplotypes were able to attribute unambiguously MS allele in 84% of the cases considering all breeds together (from 50% for Selle Français alone to 100% for French Trotters alone). Longer haplotypes will be tested and influence on parentage verification will be checked.