This work introduces a probabilistic framework for fatigue assessment in cement-based materials, with emphasis on concrete, using Bayesian inference as the central analytical tool. The approach integrates the Sparks–Menzies relationship, which links fatigue lifetime to secondary strain-rate development, into a probabilistic formulation rather than a fixed analytical expression. By shifting from deterministic expressions to probability-based estimation, the model captures the natural scatter associated with material properties, environmental variability, and loading conditions, offering a more realistic prediction of fatigue performance. A strain-driven criterion previously defined in deterministic form is re-expressed here within a Bayesian structure. Instead of assigning constant coefficients, the parameters are characterized through probability density functions, obtained from experimental data sets. This enables a richer description of uncertainty and provides fatigue estimates with quantifiable confidence levels. The proposed method embeds variability directly into the predictive process, allowing early identification of critical failure indicators during fatigue development.