Koi carp (Cyprinus rubrofuscus) are an invasive fish that has caused significant ecological impacts to freshwater ecosystems around the world. Current detection methods such as observation, netting, electrofishing, and environmental DNA can be time-, labour-, and cost-intensive, especially when fish densities are low. Evaluation of water samples for invasive fish by scent detection dogs could offer a more efficient method of detection. However, the odour profiles of such samples are likely unstable, and without effective preservation from the time of collection to the time of analysis by dogs, samples could degrade, making the target scent less recognisable to dogs. The aim of this study was to determine whether refrigeration (4°C), freezing (-18°C), or the addition of potassium sorbate (1.7 mM) could be used to preserve scent in water samples assessed by dogs. An ABACADA reversal design was employed. The dogs performed a baseline evaluation (A) of unpreserved water samples from aquaria containing carp (target scent; n = 7), goldfish (Carassius auratus; non-target distractor scent; n = 5), or no fish (non-target scent; n = 5) before preservation treatments were applied to the samples. The treatment phases (B = refrigeration, C = frozen, and D = dark cupboard with potassium sorbate) involved applying a preservation method to carp, goldfish, or no-fish water samples 7 or 8 days prior to assessment by the dogs. There was weak evidence that refrigeration impaired the dogs’ detection performance. In contrast, once stability criteria had been met, detection performances on frozen or potassium sorbate preserved samples were similar to or returned back to baseline levels. These findings suggest that freezing and potassium sorbate were effective in retarding scent degradation over the storage period and have the potential to be used as preservatives for water samples awaiting assessment by scent detection dogs. If laboratory-based canine scent detection becomes an invasive fish detection method, the use of dogs could prove more time- and cost-efficient than current detection methods.