Data on search engine market share is available, but I wonder what that looks like for Lemmy users in particular, who I would assume lean more technical than the average user, so probably use DuckDuckGo and alternates more than Google.

I use a mix of DuckDuckGo and Kagi. I’ll also use ChatGPT, which can be good if you’re careful to verify the answers it gives you as a check against hallucinations. It’s useful for short, direct answers without ads or SEO bullshit.

This article on Ars (and if you’re not a subscriber, you absolutely should be, as they are the best tech journalists out there) inspired the question: https://arstechnica.com/gadgets/2023/06/google-admits-reddit-protests-make-it-harder-to-find-helpful-search-results

Fucking Reddit. Enshittification ruins everything.

    • Kir@feddit.it
      link
      fedilink
      arrow-up
      3
      ·
      1 year ago

      Why do you have to answer like that?

      You are linking a search engine based on generative AI, which is a different things than using chatgpt per sè and, as I was saying to another user, I did not know existed.

      If you don’t like my answer you can simply not comment on that. I don’t care if you agree with me or not, be polite

      • newde@beehaw.org
        link
        fedilink
        arrow-up
        3
        ·
        1 year ago

        Don’t worry, you are in the right: a LLM is not a search engine. You might integrate it into a search engine, but that doesn’t make it a search engine.

        I mean it’s so glaringly obvious it is not a search engine: every time you ask ChatGPT for information it will give you a disclaimer it’s database is from 2021 and prior…