But it is incredible difficult to get people to switch to a different search engine (just ask Bing).
The HN community is different than the average web browsing population, but what would it take for you to switch your search engine?
(To be fully transparent, I am building an alternative web search tool, so I have additional interest in the answers to this question in addition to it being a good conversation).
I don't really know if other search engines would improve over time as I "trained" them on the subjects that I search for (as I have unwittingly done with google).
As a recent example, Google knows full well that when I search for "hugo" that what I am after is results related to a template engine, not about a film, line of clothing, politician, or poet.
I am currently unwilling to suffer through the bad results phase in the hopes that I can train the search engine up. Even then, once I have trained the search engine on my profile, what will I have gained other than teaching some other aggregator about the things I like?
I wonder if it would make sense to have a few "intent disambiguation" radio buttons to pre-prioritize results. Something like this:
"Purchasing Intent" => comparison shopping sites, retailers "Informational Intent" => Product info pages, review sites "Support Intent" => Forums, blog posts, and Stack Overflow sort of sites
Here's my favourites, some of these actually use Google: https://www.qwant.com/ https://www.startpage.com/ https://www.ecosia.org/
Once upon a time if you wrapped something in quotes, Google would search for exactly what you asked for. That hasn't been the case for several years.
I use Google for security research. It tries so hard to protect users that its ruining its search results for anything technical.
I want to be able to use URL query strings to specify stuff.
I don't want intrusive CSS/JavaScripts. I have my own settings for fonts, window width, colours, etc, and don't need yours.
I want to be able to tell it to randomize result orders.
I don't want the search to index text which is only available when JavaScript is enabled, try to guess text from pictures, try to follow form submissions at all, etc.
I want to be able to enter advanced search queries, including based on parts of the URL, based on keywords, both exact and approximate matching, and to be able to exclude based on the keywords, based on protocols, on what features the web page may include, MIME types, etc.
I don't want it to try to guess what I mean and correct it; it should instead search for what I type in.
I should hope for good results, when they are available. Options to customize this may help, and maybe also the ability to specify submissions with keywords for feedback so that users can indicate if it is good or not, with the ability to control how well this functions (perhaps with cookies).
If it uses cookies, document what each one does, in order that the user may edit the cookies using their cookies manager to the settings they want. But, allow it to work with cookies read-only or disabled, too.
The search engine should work as well in Lynx as it does in Firefox and other browsers.
It should not be HTTPS-only (although HTTPS is good to have, it shouldn't be mandatory).
It should abide by robots.txt and not index that which is disallowed by the robots.txt.
The user agent string and other request headers it uses should be documented.
I do want it to index plain text files too, and not only HTML.
Another good thing to have would be the ability to set cookies to cause it to permanently exclude certain things by default, if the user wants it (see above about documenting cookies).
I want to be able to get the results in plain text format (and possibly also JSON, CSV/TSV, etc) if the Accept header specifies such a thing.
I want the same search result as someone on the other side of the Atlantic, with a completely different browser history.