Someone (average smartphone user) complained to me the other day that there isn’t a clear solution to navigating the web because it’s either ai slop or the results are poor.

And I kinda agree, nowadays you do have to put in some more effort to get what you’re looking for.

And so I told them I’d compile a list of which search engines are best for what purpose.

So far I have:

Startpage: works, but does a poor job of finding product codes and whatnot.

Duckduckgo: gives more general results afaik

Ecosia: works great but relies on bing

I have tried things like qwant ect but haven’t really settled on those.

Any other suggestions?

I’d tell em to use udm14 but it’s still google so I’d rather keep that as a plan B.

  • kuneho@lemmy.world
    3·
    1 day ago

    I use duckduckgo for every of my web searching needs. 85% of the times I get relevant or somewhat relevant results. The rest… well, I acknowledge that I may search for super nieche things and move on.

    Only thing I don’t kinda like in duckduckgo though is that the result list may change if I navigate to somewhere and go back, but I can live with it.

  • FoundFootFootage78@lemmy.mlEnglish
    7·
    2 days ago

    Ultimately duckduckgo is the best.

    Bing and Google are out for obvious reasons. Startpage kept asking me to prove I’m human. Qwant blocked me whenever I enabled a VPN with an exit in a region they couldn’t run ads to. Ecosia is a non-profit so I’d feel bad blocking ads, but I don’t want to see ads. Mojeek returned garbage results. Kagi is ridiculously overpriced and restrictive. Brave Search often didn’t return the results I was looking for. And SearXNG I don’t remember what the problem was.

  • Matt@lemmy.ml
    5·
    2 days ago

    Running all my queries through SearXNG in normal windows and DuckDuckGo in private ones.

    • Ghoelian@piefed.socialEnglish
      14·
      2 days ago

      Same here. It’s great easily being able to block those shitty github/stackoverflow scrapers like codegrepper, answeroverflow etc.

    • mnemonicmonkeys@sh.itjust.worksEnglish
      103·
      2 days ago

      and you can downrank/ban websites.

      Holy shit, I’ve been wanting that for years. I’ll have to check them out.

      Edit: Aaaand they have a built in AI despite calling themselves “human-focussed”. That’s a turnoff. If I’m paying for a search engine, I’d prefer to have a subscription option with the AI stripped out so I’m not subsidizing it for others.

      Plus, I’m thinking about it I’m not exactly on board with needing an account with a search engine. That makes it super easy to track you. Meanwhile, using noai.duckduckgo anonymously via a VPN is way more private

      • graynk@discuss.tchncs.de
        72·
        2 days ago

        I’m not exactly on board with needing an account with a search engine. That makes it super easy to track you.

        https://help.kagi.com/kagi/privacy/privacy-pass.html

        they have a built in AI despite calling themselves “human-focussed”

        Because it’s a tool that can serve humans. Not strictly search-related, but their AI-based translation tool is genuinely great and very useful to me.

        https://help.kagi.com/kagi/why-kagi/ai-philosophy.html

        You don’t have to use built-in AI, it only triggers when you explicitly ask for it and you can completely disable it in the settings.

        They also have anti-slop filters, BTW: https://help.kagi.com/kagi/features/slopstop.html

        Finally, not sure what models the search summarizer and the translation tool use, but the assistant lets you use open weights model like Kimi, GLM, DeepSeek et al.

        • mnemonicmonkeys@sh.itjust.worksEnglish
          38·
          2 days ago

          You don’t have to use built-in AI

          I don’t care. It’s bundled in with even the basic plan, which means you’re paying for it, even if you don’t use it. And no, it’s not magically better to use open models.

          I don’t get how this concept is so hard to understand.

            • mnemonicmonkeys@sh.itjust.worksEnglish
              38·
              2 days ago

              1: I clearly established a boundary and you insisted on pushing it. Learn to take a fucking hint.

              2: You know what’s even better? Not using AI at all.

                • mnemonicmonkeys@sh.itjust.worksEnglish
                  25·
                  2 days ago

                  I feel bad for anyone you’ve ever had a relationship with. You act like an abuser.

                  Edit: Seriously? You downvoted all of my comments? You just keep getting more petty

      • ApplyingAutomation@lemmy.world
        31·
        2 days ago

        In my experience, Kagi doesn’t force AI down your throat. It only triggers an AI summary if you end your query with a “?”, and it does seem to be summarization of the search results.

        Also, as per Kagi’s pricing page, there is a higher tier for people who want more AI features as you are suggesting.

        All that being said, I use a Searxng instance 😅

        • mnemonicmonkeys@sh.itjust.worksEnglish
          18·
          2 days ago

          I don’t even want the fucking AI summaries. I don’t get how that’s so hard to understand.

          The fact the AI is in even the basic plan means you’re paying for it, regardless of whether you use it. I don’t want AI, and I especially don’t want to give money to one.

  • stellargmite@lemmy.world
    321·
    3 days ago

    Kagi. Not cheap but I seriously cant go back. Its really good for what I need - actual results at the top of the first page, based purely on what I searched for. What a concept. My needs are basic to be clear.

    • Chat_mots@jlai.lu
      6·
      2 days ago

      Searxng with custom engine weights and a proper filter of upranks/downranks domains give me as good results as kagi !

        • Chat_mots@jlai.luEnglish
          5·
          1 day ago

          No there is none unfortunately, but i’m in the process of making one! For now my settings are these. And i made a quick tool to compare results from a searxng instance to a list of desired results, you can find it here. If you want help with that, i’d be glad to help !! :)

          EDIT: more details in my settings example

    • myrmidex@belgae.social
      19·
      3 days ago

      I love how it allows me to move domains up or down. This allows me to have wikipedia results at the very top and have all X, FB, MS domains blocked from appearing in the results.

  • custom_username@reddthat.com
    5·
    2 days ago

    I often add the word “forum” to my searches to help get some actual human discussion instead of ad farming AI slop articles. Seems like the only way to figure things out these days.

    • Bronzor@lemmy.ml
      1·
      2 days ago

      Is your instance public? Wouldn’t it need to be to anonymize or privatize your usage?

  • SuspiciousCarrot78@aussie.zone
    9·
    2 days ago

    Well, at risk of downvotes…

    I know what your asking but tbh I generally don’t search, at least not like we used to. It’s weird to say that, like I’m making some claim to heightened purity or some bullshit but it’s not that at all. Search is just … noisy AF these days.

    I actually find myself using fewer sources, more directly and own my terms.

    Eg: For something like PubMed, I go to pubmed (or define a proper boolean PICO search in an indexing tool). Sci-hub is a good source too.

    https://sci-hub.st/

    Or I can hit the wikipedia API (or even better, self hosted kiwix) and find what I need.

    https://kiwix.org/en/

    Or, for known, commonly visited sites (like fan wikis), I use self-written tools.

    There are plenty of sites have public facing, nicely formatted, JSON friendly results. Others are just clean to scrape. A recent discovery was Muvitimes (for getting movie times listings).

    https://muvitimes.com/

    These tools query my white listed sites to a particular depth, collect the results I need, and present them for review.

    You can then cache the info locally to query against using something like Meilisearch.

    https://github.com/meilisearch/meilisearch

    There are other options too. Something like Scrapy can collect structured data and a Trafilatura-like extractor can remove menus, adverts, and repeated page content.

    https://github.com/scrapy/scrapy

    I’ve even use something like self hosted Perplexica. It can query SearXNG, Tavily, Exa or other sources.

    https://github.com/kiranz/perplexica

    Honestly, if it’s a quick throw away search, I’ll call my self hosted llm, get it to pull from Tavilly and give me the summary and citations. That’s been miles better than most other options for me but YMMV.

    Finally, if you’re talking about random occasional searches on phone, when I can’t access my LAN, I don’t mind DDG-lite

    https://lite.duckduckgo.com/lite