This paper presents a novel quantification of the fraction of broken waves (Q<sub>b</sub>) in natural surf zones using data from seven microtidal, wave‐dominated, sandy Australian beaches. Q<sub>b</sub> is a critical, but rarely quantified, parameter for parametric surf zone energy dissipation models, which are commonly used as coastal management tools. Here, Q<sub>b</sub> is quantified using a combination of remote sensing and in situ data. These data and machine learning techniques enable quantification of Q<sub>b</sub> for a substantial data set (>330,000 waves). The results show that Q<sub>b</sub> is a highly variable parameter with a high degree of interbeach and intrabeach variability. Such variability could be correlated to environmental parameters: tidal variations correlated with changes in Q<sub>b</sub> of up to 70% for a given local water depth (h) on a low tide terrace beach, and increased infragravity relative to sea‐swell energy correlated to lower values of Q<sub>b</sub> at the surf‐swash boundary. Q<sub>b</sub> also correlates well with the Australian beach morphodynamic model: For more dissipative beaches Q<sub>b</sub> increases rapidly in the outer surf zone, whereas for more reflective beaches Q<sub>b</sub> increases slowly throughout the surf zone. Finally, when comparing data to existing models, three commonly used theoretical formulations for Q<sub>b</sub> are observed to be poor predictors with errors of the order of 40%. Existing theoretical Q<sub>b</sub> models are shown to improve (revised errors of the order of 10%) if the Rayleigh probability distribution that describes the wave height is in these models is replaced by the Weibull distribution.