Abstract For the 2026 International Forum on Artistic Research, I will present a poster that describes an artistic reinterpration of a machine learning algorithm that engages with cultural and ideological framing of landscape photography in smartphone camera systems. ‘Land/scraping’ is a custom-designed generative image algorithm that explores how the concept of landscape is developed within generative image optimisation processes. In 2021, Google introduced Magic Eraser to its Pixel line of smartphones, enabling users to alter images with machine learning. With generative AI, editing possibilities include swapping clouds for a clear blue sky or a night sky, repositioning objects and people, or removing them altogether. To achieve this, a machine learning model infers what is in an image, what is behind an object when it is removed, or what the user would prefer the image to look like, i.e., a ‘better’ sky or a ‘better’ composition. Land/scraping questions how user experience is mediated, or ‘trained’ through generative image optimisation technology. To experiment with the boundary between algorithmic and cultural training, Land/scraping restages generative AI optimisation tools through training an algorithm to optimise according to contrasting ‘cultural’ understandings of landscapes. By presenting this research as a process diagram, Land/scraping makes visible the influence of ‘cultural atoms’ on conceptions of landscapes and proposes alternative optimisation practices. To achieve this comparison and tension, six ‘alternative image datasets’ of landscapes were created for this art experiment from different cultural, commercial, and time-specific contexts. The diagram provides an overview of how image data was processed in this art experiment, which aims to open up space for speculation on alternative forms of optimisation that can either disrupt dominant conceptions of landscape or reinforce cross-cultural or personal visual experiences through generative AI systems.