Fertility of hybrid tea roses is often reduced due to their interspesific origin but also to intensive inbreeding. New genotypes used as pollen donors represent an economic risk for a breeding program, as their influence on seed production is unknown. In this study 9 garden rose genotypes were selected from a company database as high fertile or low fertile male parents, according to the number of seeds per hybridization. Pollen morphology and in vitro germination of the selected genotypes were characterized. Pollen was either small (mean diameter = 30 urn), shrunken, and irregular (abnormal), or large (mean diameter = 30 urn), elliptical and crossed by furrows (normal). High correlations were found between the number of seeds produced per hybridization and the pollen diameter (r = 0.94) or the percentage of normal pollen (r = 0.96). In order to evaluate the predictive power of the models, we conducted regression analyses and performed a validation experiment on genotypes not present in the database and without background information on fertility. Pollen diameter and percentage of normal pollen were characterized and fitted in the regression models for seed set predictions. Validation with an independent dataset gave a good prediction for 83.3% of the data. This indicates that using either the mean pollen diameter or the percentage of normal pollen resulted in effective fertility prediction. Moreover cluster analysis of the data classified all the cultivars into various groups with varying fertility. This tool could enhance the genetic variability in crossings between hybrid tea roses, thus creating possibilities for less economically risky exploitation of new tetraploid genotypes as male parents.