I have recently started using Mojeek more regularly, and I really appreciate its focus on privacy and independent indexing. That being said, I am curious how Mojeek approaches the balance between user privacy and delivering personalized or relevant search results.
Most major search engines rely heavily on user tracking and data profiling to refine their results, which is something I am trying to move away from. Although, I do miss some of the relevancy tweaks that those engines provide like understanding my past search behaviour or tailoring results to my interests.
So I wanted to ask: how does Mojeek aim to keep search results relevant without tracking users? Are there any plans to introduce optional personalization features that still respect privacy like local device-based learning or opt-in profiles?
I believe this is one of Mojeek’s core values, any monetisation would be based on your search query, not stuff you’ve previous searched for (or based on 3rd party tracking across the web) - and this would be symbiotic enough for a search engine to finance itself.
personalization features
Are almost always based on tracking/search history. I’ve found that it’s marginally useful- e.g. I search for [famous footballer who played for X team] and then search for another [famous footballer] and search history will lean you towards someone who played for X team if it’s an ambiguous name. Maybe a famous example is Brian Cox. A lead on query may be about actors or physicists. With the power of AI nowadays, I’d take a leap of faith that you could be offered both options or a pretty good guess without knowing your history (beyond your current search) - unless the original query was brian cox, the emphasis is now on the searcher (and maybe the ui to disambiguate)
To be fair to the UI there’s only so much space, but the status quo seems to be lazy searcher.
You can personalize your search results using Mojeek Focus:
You can also share your templates and browse others’ shared Mojeek Focus templates on GitHub:
For advanced business users, you can subscribe to the Mojeek API and combine that with a local LLM. That would allow you to further personalize and refine your searching over time using personal data. For example:
Mojeek also makes it easy to send your search results to different search engines in case you’re not happy with the results you got from Mojeek.
You can use Mojeek Search Choices which is a preference that adds additional search buttons at the bottom of each search results page. With Search Choices enabled, you can send your results to a different search engine by clicking on one of the buttons at the bottom of the search results:
It is also possible to customize the search engines available in Firefox or Chrome and their derivative web browsers.
For example, in Firefox, you can set a preference to: 1) add a search box to the Firefox UI which is separate from the address bar, and 2) keep your keywords between searches. That way you can use the browser to send your keywords to a different search engine.
The thread topic is not directly related, but this post links to the instructions for adding search engines to a web browser:
If you want to improve Mojeek’s search results, the most relevant way is by giving feedback on your search results.
The feedback button appears on the bottom right of the search results page:
@mike and @ricardo81 have made some excellent points, and @mike has providers useful pointers.
On personalisation this blog post: About Ranking on Mojeek | Mojeek Blog discusses our views. Obviously our stance enables full privacy but it also has other important benefits:
Ensures neutrality: Results aren’t shaped by personal data or commercial influence. Everyone sees the same search results (given the same timing location and language settings), reducing bias and removing influence.
Prevents filter bubbles: Users aren’t trapped in echo chambers shaped by their history.
Promotes fairness: Information is ranked without favouring popular or high-traffic sites. This helps all websites get discovered, not just the already-popular ones.
You are not the product: We are not collecting, passing on or monetising user data.
Local device-based learning, or even opt-in profiles for those who want it, are certainly candidates for future development. In the meanwhile there is a lot that we are working on, and planning, that will improve search relevance.