The performance curves of electrical submersible pumps (ESPs) are commonly characterised using water as the working fluid by their manufacturers. However, ESPs are widely deployed for oil production where the fluid viscosity is significantly higher than water, and conventional methods of head estimation can become unreliable. Analytical methods rely on complex derivations, empirical methods may have limited applicability, and computational fluid dynamics (CFD) requires heavy computational resources. This paper explores the applications of artificial intelligence (AI), specifically artificial neural networks (ANNs) to develop fast and accurate models for predicting head of an ESP under varying viscosities. Two ANN architectures,a multilayer perceptron (MLP) and a fully connected cascade (FCC) network, were developed and trained on experimental data provided by (Shi et al. in Proc Inst Mech Eng, Part A: J Power Energy 235:1976–1991, 2021 [25]). The experimental data consisted combinations of various motor speeds, viscosities and flowrates with their corresponding head. The methodology included data preprocessing, development of preliminary models and hyperparameter optimisation using random search and Bayesian optimisation. Both models demonstrated strong prediction metrics, with a mean absolute error (MAE) below 0.6 ft and mean absolute percentage errors under 8% on the unseen testing dataset. Whilst the FCC model outperformed the MLP model in terms of raw metrics, the MLP model demonstrated smoother and more traditionally expected head curves. The MLP model also requires fewer number of parameters making them easier to train and more suitable for deployments. The study however uncovered an unexplained sensitivity of the MLP model to the normalisation technique used for the output data—suggesting an area for future research. Overall, this study highlights the potential of AI-based models in head predictions for ESPs.