Google is such a strong brand on the internet that some people think it is the home page of the entire web. They type facebook.com into the search field and click the first result to visit the service. I haven’t used Google Search for years because Google tracks and collects private data about its users, although I admit it is the gold standard in search. Now, an independent study has discovered a search engine that is better than Google – and without any collection of sensitive private data.
LibreTechTips conducted quite an extensive study on 11 search engines that are readily available on the net. The motivation was to compare search engines’ capability to deliver desired results. In practice, it means competing against Google. Some of the tested services, like Bing and DuckDuckGo are familiar to most people, whereas names like Mojeek or Metager may not be widely recognised yet.
The method for ranking the search engines was simple. Throw in many types of searches to the search engines, and conclude which one provides the most relevant and accurate results.
As expected, Google was a top performer, but not the best. Searx, an open source, privacy-respecting search engine was even better than Google Search. Here is the curious thing: Searx doesn’t maintain its own index, but utilizes results Google and other search engines produce.
The two best search engines were in a category of their own:
The second tier ranked way behind the top engines, but did quite well, nonetheless:
The rest of the search engines missed too many test searches:
You can read detailed descriptions of the test results and conclusions at LibreTechTips.
Searx is not an ordinary engine that collects a vast index of web site links from the internet, but a metadata ranking system. It lets engines like Google and Bing to conduct the actual search, and then it matches their results with the query.
Searx software is open source code that anyone can download, and run on their own servers. That’s ultimate privacy. In practice, of course, the vast majority of people access public Searx engines that organizations, enterprises and individuals are already running. If you trust a Searx instance operator, access the service, if not, go somewhere else.
Searx in numbers:
- has 121 contributors
- consists of 25,029 lines of code
- is mostly written in Python
- took about 6 years to create (COCOMO model)
- was started in October, 2013