I posted this here because the shower thought was “if I google something that uses reddit as a source, doesn’t that lose traffic for reddit?” But then I thought deeper and was wondering if it would send a bit to reddit and find out, or if it already had and it’s giving me a review of something the AI already had and was just giving me a link to the source?
Maybe I should’ve posted in ELI5. I dunno, if this doesn’t belong here just let me know or delete it.

I’m learning a lot right now, thanks. I’m not at all a fan of AI but sometimes it helps with searches and seems more useful, especially for the price (economically, socially and politically) but I wasn’t aware there were differences between an AI search and an LLM.
It’s not a bad way to scour the web TBH. Treat it like Wikipedia: use it for quick reference, but check its sources for details you need verified. Assume anything the LLM says directly could be completely made up.
And yes, all an LLM does is take a block of text. That’s it. It has no memory, it’s not hooked into anything directly. All the users facing systems you see are just fancy ways of generating a prompt for the LLM to read, kinda like if you pasted entire webpages into a chatbot bar.
…And on the moral aspect, I’d recommend using Z.ai, or maybe Xiaomi MiMo or Kimi, over Google for research. All their models are open weights, and trained on peanuts compared to US models, and take a tiny fraction of the resources to run.
Mistral is okay if you want a European company instead.
Or use Google LLMs directly, instead of through search: https://aistudio.google.com/
Google is still Google, but they are a bit less bad than Anthropic/OpenAI because they use TPUs instead of GPUs, because they use a more efficient software architecture, and… well, they already had all the data, instead of having to scrape it.