Big caveat here is how people are using the LLMs. Here they were using them for things like information recall and ideation. LLMs as producer and human as editor / curator. They did not test another (my preferred) mode of LLM use - human as producer and LLM as editor / curator.

In this mode of use, you write out all your core ideas as stream of consciousness, bullet points or whatever without constraints of structure or style. Like more content than will make it into the essay. And then have the LLM summarize and clean it up.

Would be curious to see how that would play out in a study like this. I suspect that the subjects would not be able to quote verbatim, but would be able to quote all the main ideas and feel a greater sense of ownership.