How many servings of vegetables and fruit do I actually need per day?
Five servings of vegetables and fruit per day is the threshold that offers the greatest protection. More than five provides no additional benefit, but fewer demonstrably costs you health gains. Potatoes and fruit juice do not count.
Five servings of vegetables and fruit per day is sufficient, and eating more offers no additional benefit. Going from two to five servings is associated with a 13% lower risk of dying during the study period, on average. Above five, the effect levels off. This finding comes from a meta-analysis of 26 large population studies involving nearly 1.9 million participants and more than 145,000 deaths1. This is a statistical association, not a proven cause, but the size and consistency of the evidence are impressive.
The strongest effect is seen for lung disease: a 35% lower risk at five servings compared with two. For cardiovascular disease, the reduction is 12%. Plant-based compounds with anti-inflammatory and antioxidant properties are considered a possible explanation2. How large their contribution actually is remains unknown.
Not everything counts equally. Fruit juice and potatoes were not associated with lower mortality in the research. Starchy vegetables such as peas and corn also showed no clear benefit1. Leafy greens, legumes, non-starchy vegetables and whole fruit perform best.
Most people do not reach five servings. In the United States, fewer than one in eight people eat enough fruit, and fewer than one in ten eat enough vegetables3,4. Notably, more than half of the people who eat too few vegetables believe they are doing just fine5. Tracking what you eat, through an app or a food diary, works better than relying on your own perception.
All claims are based on observational population data and one small randomised study. The meta-analysis (PMID 33641343) is large and consistent, but the evidence remains associative. The claim about plant-based compounds (PMID 31521398) is biologically plausible, but has not been experimentally quantified in the source used. The small study in gestational diabetes (PMID 37630814) is too limited for broad conclusions.