Variational quantum algorithms are highly sensitive to ansatz design, data encoding, and optimizer hyperparameters, yet most quantum machine learning (QML) studies still tune these choices by hand. Automated hyperparameter optimization (HPO) and automated QML (AutoQML) address this gap by extending the classical Combined Algorithm Selection and Hyperparameter optimization (CASH) problem into the quantum domain, but the resulting literature is fragmented across search strategies and at least three unrelated systems that share the name "AutoQML." This paper surveys automated optimization techniques for QML, organizing the literature into a taxonomy of search strategies, architecture search, model-family selection, multi-objective formulations, and explainability integration, and compares representative open-source systems — AQMLator, the Fraunhofer IPA AutoQML framework, and our own Quoptuna platform — across a common set of capabilities. Among the systems compared here, none combines full architecture search, hyperparameter optimization, and explainability, and explainability-integrated AutoQML remains, to our knowledge, largely unaddressed in the 2021–2026 literature we surveyed. We close by identifying open challenges — barren plateaus' effect on search signal, evaluation cost, benchmark fragmentation, and search-process explainability — toward practical, reproducible AutoQML.

quantum machine learning,AutoQML,hyperparameter optimization,variational quantum algorithms,NISQ,neural architecture search,,explainable AI