Over the past years, there has been an increasing interest in the design and evaluation of home appliances that have been increasingly interconnected and personalised, integrating visual, auditory, haptic and multimodal interfaces catering to diverse user preferences and contexts. One of the greatest challenges in these human-machine interaction environments is to assess ecological validity in the studies, i.e., approximate the testing environment to real-life situations. Although the alignment between the user’s mental models and response to the environment has been covered in the literature mostly through self-report data collection methods, there has been a lack of understanding about the procedures to undertake when assessing the user’s visual attention in such scenarios. This study explores eye-tracking analyses to assess user experience in smart home appliances by providing three groups of users (N = 30) with different instructions, including written, verbal and no instructions. Gaze data obtained during the interaction were triangulated with responses from the User Experience Questionnaire (UEQ). Results suggest that written instructions are helpful direct initial attention but can potentially lead to entanglement during control adjustments. Verbal guidance seems to promote efficient visual attention and focused task execution for simpler tasks, while written instructions better support tasks requiring detailed cognitive processing. In addition, proximity between the appliance interface and the user can compromise the natural behaviour in a smart home environment. Findings may inform a protocol for assessing smart home appliances and examine the impact of instructions.