Rod-like, flower-like and caddice clew-like zinc oxide architectures were fabricated by increasing annealing time under a simple solution route. The annealing time altered the orientation of rods, which made rods form flower. The well-defined nanoflower was composed of orderly and radical nanorods. Of all three samples, the sensor-based nanoflower showed the high response, short response/recovery time, and excellent selectivity to ethanol under the optimal working temperature of 360°C. The enhancement of ethanol-sensing properties is owing to 3D hierarchical and layer-by-layer structure which provide more gas diffusion pathways. The above results suggest that the 3D hierarchical feature with large surface-to-volume ratio is a good candidate for ethanol detection.