The aim of this study was to provide insight on the herd immunity of COVID-19 using a mathematical model taking into account different activity levels across age cohorts. When taking both age and activity level into account, the model showed that at every reproductive number (R 0 ), defined as every other infection per one infection, the disease-induced herd immunity level is lower than the classical herd immunity level, calculated by mathematical models used for vaccination impact. For example, assuming a basic R 0 = 2.5, the model estimates that herd immunity is lower (43%) than models assuming homogeneous immunization (60%). This model estimated that the group of age 13-59 with the highest activity level is more likely to have the highest herd immunity levels (above 70% fraction infected), but the young (age<5) or the elderly (age≥60) with the lowest activity level are least likely to develop herd immunity (between 15-18% infected). In addition, the model also shows that herd immunity may be dependent on the presence of preventive measures, showing higher levels than scenarios with no preventive measures.