Abstract: Conductive elastomer-based flexible strain sensors generally suffer from issues such as a narrow detection range, low sensitivity, complex fabrication processes, and high costs, which limits their application. Therefore, a simple gravity-driven method was employed to fabricate flexible strain sensors based on natural rubber (NR)/carboxylated multi-walled carbon nanotubes (c-MWCNTs) with a gradient conductive network structure. The hydrophilic properties of c-MWCNTs and the gravitational force acting on them were utilized to control the distribution of c-MWCNTs within the NR matrix, thereby forming a gradient structure within the NR matrix. This structure exhibits a continuous concentration gradient of c-MWCNTs, with concentration increasing from top to bottom, resulting in a conductive network that transitions from sparse to dense. Compared to conventional sensors prepared via blending methods with a single conductive network, the designed sensor demonstrates superior sensing performance due to the differences in elastic modulus and conductive pathways between hierarchical conductive networks. The sensors exhibit gauge factor (GF) values of 9.025 and 61.127 at strain of 0–200% and 200–240%, respectively, while offering a wide detection range (0–240%) and fast response/recovery times (164 ms/1.07 s). Furthermore, the sensor can conform to various joints of the human skin, enabling monitoring of various physiological signals, and holds significant application potential in fields such as human health monitoring and human-machine interaction.