Hydrogen’s interaction with steel pipelines poses significant challenges for infrastructure in the transition to sustainable energy systems. While hydrogen offers a clean alternative to fossil fuels, its deleterious effects on steel pipelines accelerate degradation through various mechanisms, including hydrogen-assisted fatigue crack growth. Traditional approaches for characterizing these effects are time- consuming and costly, necessitating the development of efficient predictive models. This study presents machine learning approaches for predicting hydrogen-assisted fatigue crack growth in API 5L pipeline steels X52, X70, and X100 [1]. Artificial neural network and random forest models were developed using experimental data comprising hydrogen pressure and stress intensity factor as inputs to predict fatigue crack growth rates. The dataset, sourced from controlled experiments conducted at various hydrogen pressures and consistent loading conditions, contains over 400 observations [2]. The artificial neural network models demonstrated superior predictive capability, achieving Pearson correlation coefficients values exceeding 0.92 for all steel grades. Notably, the models successfully predicted fatigue behavior at hydrogen pressures not included in the training dataset, demonstrating robust interpolation capability. In contrast, random forest models, despite achieving higher correlation during validation, exhibited limited interpolation performance for unseen pressure conditions. The trained neural network models were used to generate three-dimensional surface plots relating hydrogen pressure, stress intensity factor, and crack growth rate. Analytical polynomial equations were extracted from these models to enable practical implementation without specialized softwares. Validation against phenomenological models revealed that the machine learning approach provided superior predictions, particularly at higher pressures and in the intermediate growth-rate regime. This work demonstrates that artificial neural networks can effectively capture the complex, non-linear relationships governing hydrogen-assisted fatigue. The developed methodology provides a robust framework for evaluating pipeline integrity in hydrogen environments and directly contributes to the advancement of safer, data-driven design practices for hydrogen infrastructure.