Currently, I prefer the Kagi model as my daily driver -- pay a small fee each month and not worry about having your search data, in all its forms, sold instead. They have some neat features, as well, like AI summaries and the ability to ask questions about pages. It's those two features I have used to determine what kind of content, even musical, may be on the page for me without having to go to the page, itself, first. I can reduce page-load-waiting on my end, in other words, because I am able to make better judgement calls beyond any provided preview text / provided metadata.

There is also SearXNG, which is an open source, privacy-respecting, hackable metasearch engine. It can be self-hosted and also has a number of instances you can access.

A good amount of this is available on Mojeek: https://blog.mojeek.com/2023/08/mojeek-operators-a-guide.html (my employer).

Search is keyword based so the words should be matched in most cases without quotes (omitting stemming here as a possibility); you cannot -"two words" but you could -blind -desire

Swisscows and Yep both seem to work well.

Why not searx

I like to use https://searx.neocities.org/ It randomises your search across several searx instances. It's not always reliable as some instances are sometimes down or exceed request limit. But retrying often works. Worst case i use mojeek or ddg

afaik duckduckgo knows all of this.

site: limits the site to search on, even shows a warning for that.
"quotes" make sure all results contain the word.
-apple make sure you wont see results where the page contains this word

Been using qwant recently and have to say, its pretty nice so far