When determining an animal’s food preference based on comparative consumption, a major problem is the potential for individuals to over-eat, rendering subjects unavailable for subsequent tests as well as exposing them to potentially adverse health implications. Here, we explored alternative, simple and quick ways of assessing food preference that involved only minimal training and avoided satiation. In Study 1, we investigated whether behaviour directed towards inaccessible food predicted consummatory choice. Following a prime with small quantities of two types of food of putatively different quality, 18 pet dogs were concurrently presented with the same two food types in a manner that they could see and smell, but not physically access. Time spent investigating the two inaccessible food types was measured. Subsequently, the dogs were given the opportunity to consume the same two food types but this time in a restricted-intake consummatory test: two puzzle feeders were presented simultaneously, each containing 18 pieces of one of the two food types. The proportion of food types making up the first 18 pieces consumed was recorded. Subjects showed a highly significant preference for the food type predicted to be of higher value in both the non-consummatory and the consummatory test. Moreover the proportion of time spent investigating the two food types in the non-consummatory test was very similar to the proportion of the two food types consumed from the puzzle feeders, demonstrating validity of both test methods. There are, however, some limitations to the consummatory test, in particular that it can only be valid if the two food types to be compared are of similar ease to extract from the boards. To assess the generalizability of the non-consummatory food preference test, in Study 2 we investigated test-retest reliability in 20 dogs, as well as consistency of preferences across populations by comparing 20 owned pet dogs and 25 dogs housed in rescue shelters. Findings were consistent across repetitions of the test as well as across the two test populations. Thus, relative food preference could be confidently and reliably inferred from behaviour during a non-consummatory exposure. The ease and speed of use, the lack of training required and the avoidance of satiation make this an ideal test for determining food preference.