> 0% of LLM users could produce a correct quote, while most Brain-only and Search users could

I think a better interpretation would be to say that LLMs gives people the ability to "filter out" certain tasks in our brains. Maybe a good parallel would be to point out that some drivers are able to drive long distances on what is essentially an "auto-pilot". When this happens they are able to drive correctly but don't really register every single action they've taken during the process.

In this study you are asking for information that is irrelevant (to the participant). So, I think it is expected that people would filter it out if given the chance.

[edit] Forgot to link the related xkcd: https://xkcd.com/1414/

I think the “crushing nihilism” pro-AI argument is what makes me most depressed. We are going to have so much fun when we do not communicate with other humans because it is a task that we can easily “filter out.”