Innovative methods to collect dietary data at multiple times across the year are needed to better understand seasonal or temporal changes in household diets and measure the impact of nutrition-sensitive agricultural programmes in low-income countries. The present study aims to validate a picture-based research tool for participants to self-record their household’s dietary diversity each month in villages of Manyoni District, Tanzania. Pictorial record charts were developed to reflect local food resources. In 113 randomly selected households, the person responsible for food preparation was trained to mark all items consumed by any household member within the home, or prepared for consumption outside the home, for a single recording day. The next day, an interview-based household 24-h food recall (H24HR) was collected for the same period. Separate analyses tested agreement (a) between picture charts and H24HR and (b) between H24HR following chart completion and on an alternative day. Concordance between methods differed between food groups and items but was high to very high for all cereals, vegetables, pulses, legumes and nuts and almost all fruits. Recording of ten items (including non-cultivated fruits and ingredients of mixed dishes) differed significantly between H24HR assessments, all of which were reported by more households in interviews following chart completion. Results suggest potential for visual prompts and the contemporaneous nature of data collection to improve the accuracy of interview-based recall. With adequate investment in developing and implementing context-adapted tools, pictorial charts may also offer an effective standalone method for use at multiple time-points in agricultural programmes.