Path planning is a core enabler of safe and efficient autonomous flight for unmanned aerial vehicles (UAVs), yet selecting a suitable planner remains challenging because performance depends jointly on workspace dimension (2D or 3D), obstacle dynamics (static or moving), and the computing and timing constraints of the platform. This review aims to support planner selection by assessing each algorithm family separately within each environment type. Following a comprehensive literature search, a systematic review identified 128 peer-reviewed Q1/Q2 journal articles published between 2022 and 2026, of which 88 studies met the inclusion criteria and form the core evidence base. A common mathematical formulation of UAV path planning is established, identifying shared objectives and constraints as well as scenario-specific requirements, including mission-level location selection, scheduling, and game-theoretic variants. The methods are organised into six paradigm classes (classical, sampling-based, optimisation-based, metaheuristic, learning-based, and hybrid) and analysed across four environmental quadrants defined by 2D/3D and static/dynamic conditions, referred to as a dimension-aware classification. Study-level evidence is stratified by platform and validation level, and no pooled performance score is computed because hardware, experimental environments, and evaluation protocols vary substantially across studies. The review also defines a six-stage validation ladder from model-in-the-loop testing to flight trials and discusses safety, uncertainty, and explainability in learning-based planners, while relating symmetric and asymmetric problem structures to the selection of appropriate planning algorithms. Overall, the findings indicate that planner maturity should be judged by deployment readiness, including onboard timing and validation beyond simulation, rather than by simulated success rates alone.