Addressing the disconnection between routing and stowage planning decisions and the lack of carbon emission constraints in Ro-Ro vehicle river–sea intermodal transport, this study investigates the integrated optimization of routing and stowage planning under carbon tax policies. A mixed-integer nonlinear programming model is developed, considering the characteristics of the river–sea intermodal network, vessel stowage safety, and carbon emissions. To solve the model, an Improved Adaptive Large Neighborhood Search (IALNS) algorithm is proposed. By embedding business processes and allocation logic into a multi-stage greedy construction mechanism to generate high-quality initial solutions, and by designing dedicated destroy and repair operators, the algorithm’s performance is significantly enhanced. Two sets of numerical experiments are conducted to validate the model and algorithm: one based on realistic transport instances along major ports in the Bohai Rim and Yangtze River regions of China, and another using multi-scale random test instances. Results indicate that the IALNS outperforms Gurobi, as well as conventional adaptive large neighborhood search, genetic algorithm, and differential evolution algorithm, achieving average objective improvements of 10.43%, 11.83%, and 27.37%, respectively. Fuel price is identified as the dominant factor influencing total cost. The carbon tax policy exhibits a significant diminishing marginal effect on emission reduction: a too-low tax rate fails to effectively incentivize emission reductions, whereas a too-high rate leads to saturated emission reduction effects while carbon-related costs continue to rise.