I Quit Google Search for AI–and I'm Not Going Back

I'm wondering why, in this day and age, it is still difficult to search for a specific type of jeans, with a specific size, style and color, and availability. And filter it by shipping method/cost.

Same for other products. E.g. resistor 0805, 1/8W or better, etc. etc.

Why does every online shop have their own (often broken) filtering system? Why can't the information be gathered by an AI, and turned into a super filter?

I don't think I am there yet. I got too confident about AI and was lazy to do some search instead or think through the problem solution. I wasted my 3 hours by not doing so. LLMs are getting better but search and stackoverflow still helps.

Every time I try an AI search engine, I always end up going back to Google. Their AI summaries at the top of search results are, honestly, "good enough."

This is, in part, because most of my Google searches are 1-4 words; as opposed to the full prose that I need to type into an LLM.

What am I missing?

I find LLMs are useful if I don't know enough to know what to search for; if I need to craft a "long-form" search query; or otherwise need to learn something in a Q-and-A format.

Edit: One of the things that ChatGPT has replaced Google, for me, is when I need a quick joke image. Instead of searching through Google Images, I just think of a quick prompt of something silly. I'm using some silly ChatGPT images in some test scripts; and now that the US courts declared they aren't protected by copyright, I know there is 0% chance I will get in trouble for them.

People quit Google Search to avoid "search-optimized junk" just to use AIs that answer using "Search" lol. Look closer at the sources your AI answers give you. They're not always pretty.

I think code monkeys or journalists etc. have GREAT experience with LLMs.

In case of niche and more technical knowledge demanding task, LLM just fails. Eg. in my case, reverse engineering. It just spits BS all day long.

I find this a little worrying from an ethical point of view. The suggestive power of web-based LLMs could be far greater than that of traditional search, and this could well be commercially exploited if we're not careful. With traditional search, providers like Google throw in ads that look like results and bias the genuine results based on funding. This is dark, but can be circumvented with some basic guidance on how to interpret the UI, and the web pages themselves are not affected. LLMs provide a wall of text which all appears reasonable, but could very well be biased or manipulative in highly subtle ways based on whichever political parties or corporations have paid the most for influence. Whilst in principle you can tell users to externally verify the claims, people will bother less and those external sources will become less accessible/useable as LLMs become more widely used. I hold out hope that they can be mitigated, but I don't yet see how.

Its early days but for our B2B SAAS, we are seeing an uptick in traffic from Chatgpt and other similar AI tools. I even have prospects sharimg their prompts as i always ask them. Google search is going to have serious issues in next few years if they continue on their current path of literring page 1 with garbage "sponsored posts".

https://archive.ph/3EqIe

This article uses an LLM to summarize itself into three key points. Those three points are milquetoast, making me much less interested in getting behind the paywall and reading the actual article.

Maybe this is a value add. Maybe the article is milquetoast and should be skipped. But either way it’s an example of how LLMs are not just competing with search they’re degrading the reading behaviors which make search more valuable (details, specifics, exact match).

It is an underrated point though that LLMs are competitive with Google search because they remove anti-user ads and SEO spam. In the long run LLMs will have that same problem though, because it’s a business dynamic not a technological dynamic.

I've been kind of expecting to see the LLM impact on Google's 10-Q for a while now.

Even if LLM queries are just a small fraction of overall internet queries, I suppose they're stealing eyeballs from higher-value searchers.

For "search engine" tasks I have switched to Kagi.com and find it more useful than Google's results. I am not using it for shopping but to find original content that offers multiple perspectives on topics I am researching. I find Perplexity helpful when I am trying to get a basic overview because it provides links to substantiate the content. Your mileage may vary.

Microsoft is barely even mentioned - a major defeat for Satya. During its entire existence Bing has been waiting for an inflection point. The inflection point came and Bing failed to take advantage of it. Perplexity is what the new Bing should have been. At this point it is fair to say Microsoft has squandered their AI advantage they got through early investment in OpenAI.