In this study, we delve into the generation of novel artificial knotted proteins, leveraging state-of-the-art computational techniques such as EvoDiff and RFdiffusion, in tandem with ProteinMPNN. Our aim is to broaden the spectrum of existing protein structures with novel knotted configurations, thereby deepening our insight into the intricate phenomenon of protein knotting. Our findings reveal that the generated artificial proteins closely mimic the natural occurrence of knotted proteins, with a comparable percentage exhibiting non-trivial topologies. Additionally, we introduce several knot types previously unobserved in natural proteins. At the heart of our study is the curated dataset of these artificial knotted proteins, aligned with their natural counterparts for comprehensive comparison. This dataset can serve as a benchmark, encouraging the development and application of new protein generation methodologies.