Cognitive bias testing measures how emotional states can affect cognitive processes, often described using the “glass half-full/half-empty” paradigm. Classical or operant conditioning is used to measure responses to ambiguous cues, and it has been reported across many species and contexts that an animal’s cognitive bias can be directly linked to welfare state, e.g., those in better welfare make more optimistic judgements. Cognitive bias testing has only recently been applied to animals and represents a key milestone in welfare science: it is currently one of the only accurate methods available to measure welfare. The tests have been conducted on many farm, laboratory, and companion animal species, but have only been carried out in zoo settings a handful of times. The aims of this review are to evaluate the feasibility of cognitive bias testing in zoos and its potential as a tool for studying zoo animal welfare. The few existing zoo cognitive bias studies are reviewed, as well as those conducted on similar, non-domesticated species. This work is then used to discuss how tests could be successfully designed and executed in zoo settings, which types of tests are most appropriate in different contexts, and how the data could be used to improve animal welfare. The review closely examines the many variables are present in the zoo which cannot be controlled as in other settings, termed here the Zoo Environment (ZE) Variables. It is recommended that tests are developed after consideration of each of the ZE Variables, and through strong collaboration between zookeepers, managers, and academic institutions. There is much unexplored potential of cognitive bias testing in the zoo setting, not least its use in investigating animal welfare in zoos. It is hoped that this review will stimulate increased interest in this topic from zoo managers, scientists, and industry regulators alike.