Some of the examples used in the Twitter thread Paul was referring to would be better served by a manually curated directory service with a possible addition of a search engine only surfacing content from the sites in the directory.

For health information and recipes in particular there are only a handful of really high quality sites that have quality content for 95% of the information most people need. I bet if you wanted to increase the coverage to 99%, that list would expand to less than a thousand sites. At those numbers manually curating the information would be easily achievable.

How to get people to use your top notch Google replacement instead of Google, however. That's the hard problem.