Abstract Objective: To characterise dietary habits, their temporal and spatial patterns and associations with BMI in the 23andMe study population. Design: We present a large-scale cross-sectional analysis of self-reported dietary intake data derived from the web-based National Health and Nutrition Examination Survey 2009-2010 dietary screener. Survey-weighted estimates for each food item were characterised by age, sex, race/ethnicity, education and BMI. Temporal patterns were plotted over a 2-year time period, and average consumption for select food items was mapped by state. Finally, dietary intake variables were tested for association with BMI. Setting: US-based adults 20-85 years of age participating in the 23andMe research programme. Participants: Participants were 23andMe customers who consented to participate in research (n 526 774) and completed web-based surveys on demographic and dietary habits. Results: Survey-weighted estimates show very few participants met federal recommendations for fruit: 2·6 %, vegetables: 5·9 % and dairy intake: 2·8 %. Between 2017 and 2019, fruit, vegetables and milk intake frequency declined, while total dairy remained stable and added sugars increased. Seasonal patterns in reporting were most pronounced for ice cream, chocolate, fruits and vegetables. Dietary habits varied across the USA, with higher intake of sugar and energy dense foods characterising areas with higher average BMI. In multivariate-adjusted models, BMI was directly associated with the intake of processed meat, red meat, dairy and inversely associated with consumption of fruit, vegetables and whole grains. Conclusions: 23andMe research participants have created an opportunity for rapid, large-scale, real-time nutritional data collection, informing demographic, seasonal and spatial patterns with broad geographical coverage across the USA.