Luke Thorburn is a doctoral researcher in safe and trusted AI at King’s College London; Jonathan Stray is a Senior Scientist at The Center for Human-Compatible Artificial Intelligence (CHAI), Berkeley; and Priyanjana Bengani is a Senior Research Fellow at the Tow Center for Digital Journalism at Columbia University. The question of whether there is a meaningful difference between search and recommendation comes up frequently in discussions of platform harms, where recommendation is sometimes said to show people things they didn’t ask for. There are of course differences, but in many ways these types of systems differ in degree, not in kind, so it’s hard to draw a clean line between them – especially the sort of line you’d want when designing regulation. Both search and recommendation are types of ranking algorithms. These are the systems which filter and order everything in the digital world, reducing it to the scale of human consumption. In this article we’ll map out dozens of real-world ranking products and features along five axes. These are: (1) the explicitness of the ranking signals, (2) the degree of personalization, (3) the explore-exploit tradeoff, (4) the openness of the inventory, and (5) the balance between engagement and content signals. We’ll pay particular attention to the many systems that fall “in the middle” and confound clean distinctions. For instance, YouTube has a search bar and recommendations on the homepage, but it also lists related videos after the user has already clicked on something, which is somewhere between the explicit query of search and the query-free…What’s the Difference Between Search and Recommendation?