Automated grading using Generative AI tools is an ongoing research topic as such platforms brings opportunity to reduce instructor’s workload. Some questions remain about how effective those tools are and if they are truly able to replicate an instructor grading and feedback. As part of a wider research project, we are analyzing the grading output of ChatGPT4 for a database design assignment. Our initial observations show the limitations of the platform even when provided with references such as assignment solution, grading rubric and grading examples.