Arming oncolytic adenoviruses with therapeutic transgenes is a well-established strategy for multimodal tumour attack. However, this strategy sometimes leads to unexpected attenuated viral replication and a loss of oncolytic effects, preventing these viruses from reaching the clinic. Previous work has shown that altering codon usage in viral genes can hamper viral fitness. Here, we have analysed how transgene codon usage impacts viral replication and oncolytic activity. We observe that, although transgenes with optimized codons show high expression levels at the first round of infection, they impair viral fitness and are therefore not expressed in a sustained manner. Conversely, transgenes encoded by suboptimal codons do not compromise viral replication and are thus stably expressed over time, allowing a greater oncolytic activity both in vitro and in vivo. Altogether, our work shows that fine-tuning codon usage leads to a concerted optimization of transgene expression and viral replication paving the way for the rational design of more efficacious oncolytic therapies.