The clinical outcome of advanced gastrointestinal (GI) cancers (especially pancreatic and oesophageal cancers) is dismal, despite the advance of conventional therapeutic strategies. Cancer gene therapy is a category of new therapeutics, among which conditionally replicative adenovirus (CRAd) is one promising strategy to overcome existing obstacles of cancer gene therapy. Various CRAds have been developed for GI cancer treatment by taking advantage of the replication biology of adenovirus. Some CRAds have already been tested in clinical trials, but have fallen short of initial expectations. Concerns for clinical applicability include therapeutic potency, replication selectivity and interval end points in clinical trials. In addition, improvement of experimental animal models is needed for a deeper understanding of CRAd biology. Despite these obstacles, CRAds continue to be an exciting area of investigation with great potential for clinical utility. Further virological and oncological research will eventually lead to full realisation of the therapeutic potential of CRAds in the field of GI cancers.