One way to take this is that this shows that the agents can't discover anything new, another way to take this is as confirmation that the AI can correctly identify the same sites that human experts can, and just suck at searching literature.