Since emerging in 2020, clade 2.3.4.4b H5N1 high pathogenicity avian influenza virus (HPAIV) has disseminated globally, causing substantial infection and mortality in wild birds and poultry. Additionally, numerous spill-over events into mammals have occurred, including mass mortalities and sporadic human infections. Such events can drive the acquisition of mammalian-adaptive mutations, which further increase the risk to humans; the most well characterised mutation is the glutamate to lysine change at amino acid position 627 in the polymerase basic 2 protein (PB2-627K). Across avian influenza viruses, PB2-627K enhances polymerase activity and replication in mammalian cells, increasing zoonotic and pandemic potential. However, bioinformatic analyses indicate that PB2-627K remains rare among clade 2.3.4.4b H5N1 viruses circulating in avian populations. Here, we investigated the impact of PB2-627K on viral fitness in avian hosts to assess the potential for mammalian-adapted viruses to re-establish in birds, mimicking infection of avian species following scavenging on mammalian species where PB2-627K has already emerged. Using a representative H5N1 virus (European genotype AB), PB2-627K increased polymerase activity and replication in human cells without compromising replication in avian cells. A mammalian-derived PB2-627K H5N1 isolate from a fox exhibited comparable replication kinetics and transmission efficiency in chickens and ducks to a closely related PB2-627E virus. Notably, PB2-627K remained genetically stable at consensus level following infection and transmission in both avian hosts. These findings demonstrate that mammalian-adaptive mutations acquired during spill-over can be maintained in avian populations, potentially elevating zoonotic risk and underscoring the importance of surveillance for such mutations in birds.