Assembly assistance systems are widely used in industrial assembly to help manage complexity due to various product variants and to avoid quality issues. The effectiveness of these systems heavily depends on clear and precise assembly instructions. However, these instructions are often still created manually, which can lead to errors, and the setup process can be very time-consuming. To improve the setup process for informational assistance systems, this article proposes a novel system for the automated generation of assembly instructions using the concept of teaching by demonstration. The foundation of the developed system lies in the application of state trees, which interpret the assembly process captured by cameras and convert it into an organized assembly sequence. The system also includes a human–machine interface that enables additional input options for further enrichment of the assembly instructions with context-based information. The assembly instructions created can be transferred via an interface to various assembly assistance systems for assembly process execution. The developed system was evaluated by a user study, which revealed both its potential for reducing the time needed to set up assembly instructions and needs for further research.