This paper focuses on Order Acceptance and Scheduling (OAS) problems in make-to-order manufacturing systems, which handle both acceptance and sequencing decisions simultaneously to maximise the total revenue. Since OAS is a NP-hard problem, several heuristics and meta-heuristics have been proposed to find near-optimal solutions in reasonable computational times. However, previous approaches still have trouble dealing with complex cases in OAS and they often need to be manually customised to handle specific OAS problems. Developing effective and efficient heuristics for OAS is a difficult task. In order to facilitate the development process, this paper proposes a new genetic programming (GP) method to automatically generate dispatching rules to solve OAS problems. To improve the effectiveness of evolved rules, the proposed GP method incorporates stochastic behaviours into dispatching rules to help explore multiple potential solutions effectively. The experimental results show that evolved stochastic dispatching rules (SDRs) can outperform the tabu search heuristic especially customized for OAS. In addition, the evolved SDRs also show better results as compared to rules evolved by the simple GP method.