In the context of dog training it would be of utmost importance to understand the mechanisms that control how attention to humans is allocated by dogs, and how this can be modified by training. This study aimed at characterising the patterns of attention paid to their handler by dogs with different training level, while performing typical obedience tasks in conditions of increasing difficulty. The study involved 64 dogs, classified as expert (completed an obedience course of 6 months, N = 20), intermediate (in training, N = 21) and novice (no training received, N = 23). In the experiment, the handler had to make the dog perform a sequence of three exercises. As first exercise (stay), handlers had to obtain a waiting posture by the dog, step back by 2.5 m and wait 30 s. The following two exercises were identical to the stay, but respectively a bowl with some food (food) and a toy (toy), were placed close to the dog before asking the desired waiting posture. The average duration of gazes (GL) and the frequency of gaze shifts towards the handler (GF) were collected in the two minutes preceding the exercises (baseline attention) and while the dogs were performing the requested exercises. In baseline GL was affected by the dog’s sex (F_2,57 = 4.98, P = 0.011, GLM anova), being highest in intact females, and training level (F_2,57 = 5.27, P = 0.008) being highest in expert dogs. During obedience exercises GL was affected by the exercise (F_2,158 = 11.57, P < 0.001) being highest in stay, and by the training level (F_2,158 = 25.02, P < 0.001), being highest in expert and lowest in novice dogs. The frequency of gaze shifting was affected by an exercise*sex interaction (F_2,154 = 2.48, P = 0.046), with intact females showing higher GF in toy than all other exercise*sex combination. Moreover, GF was affected by the dog’s training level (F_2,154 = 6.64, P = 0.002), being highest in novice dogs. Different mechanisms could contribute to explain our results, including direct effects of training on interspecific attention patterns as well as indirect effects of the dog-owner relationship. The role and weight of such mechanisms in shaping dogs’ attention to owners’ remains to be addressed.