To address the prevalent problems of low efficiency, disconnection between“travel”and“learning”, and superfi cial learning in geography study trips, and to eff ectively foster students’geographical key competences, this paper analyzes the geography exam questions of the Guangdong College Entrance Examination from 2021 to 2025, clarifi es the requirements for students’key competences and key abilities in geography, and introduces deep learning theory to construct a four-in-one deep learning model:“exam question analysis – context creation – inquiry practice – multi-dimensional evaluation”. This model organically integrates the four core features of deep learning—active construction, critical thinking, knowledge transfer, and collaborative communication—into each link, forming a closed-loop system that promotes students’ shift from knowledge memorization to meaning construction. Based on this, a concrete study trip curriculum and evaluation system are designed with Huangshan–Xidi Village in Anhui as an example. Teaching practice proves that this model can eff ectively improve students’problem-solving abilities in complex situations, systematically cultivate their geographical key competences and key abilities, and features high feasibility and popularization value.