If I say "show me the best winter gloves, and only from sites that you can verify actually product tested" and it follows the instruction (ignoring sites that just have a list of popular search results aggregated) then it is better. If it doesn't do what I want, I expect to be able to follow up and teach it.
I expect the chat style stateful search to take instruction for what type of sites I want results from. "Return me a list of websites with recipes for Bolognese that do not have a long story above the recipe. Build a table with the top five results normalized for portion size, comparing and contrasting the ingredients. Highlight unique ingredients in bold."