This research explored the use of generative artificial intelligence to augment the training dataset of bone surface modification images for improving the classification of archaeological artifacts using computer vision. A foundational dataset of existing bone surface modification examples was leveraged to train a generative diffusion model, which was fine-tuned using Low-Rank Adaptation (LoRA). This model was then employed to generate synthetic bone surface modification images which were incorporated into a computer vision application designed to classify artifacts based on whether the bone surface modifications were associated with bones containing meat or bones without meat. The study demonstrated that synthetic datasets can significantly enhance model performance, particularly for models trained on smaller, underperforming datasets. Notably, combining the original dataset with the generated synthetic data led to improved accuracy and F1 scores, highlighting the value of generative AI in overcoming limitations of limited real-world data in archaeological applications.