In theory I understand their change to make search -> AI search instead, but I have a hard time believing that this was benchmarked in any way and not rushed to desperately remain relevant in the AI race.
Given how AIs work, if they weren't trained with last night's sports news, you can't get an accurate answer this morning. All you can get is a hallucination.
Different experience for me, IMO AI mode saved Google search. The classic search became pretty much worthless due to decades of SEO winning over good results. But AI mode made Google search far more useful for me and gives me what I'm looking for 99.9% of the time.
I do understand some people (website operators who used to rely on SEO or google stockholders after the recent earnings report) being upset about AI mode, but as a search user I find it much more effective than where search was just prior to AI mode.
Unless I'm searching for something very niche, I'm often quite surprised by how 'directionally right' Google's AI overview is, even if I retain a lot of the skepticism I had towards it.
Though, most of the time I use it to find a reference and ignore the AI summary. Or to put a label on a topic I'm pretty sure exists but don't have a name for (e.g., "<insert problem statement here>. I'm pretty sure this is a known problem in <insert domain here>. Is there a name for it and what is the most recent literature?"
> What is the equivalent focal length on APS-C of 80mm on full frame?
and get the correct answer. It's not good compared to frontier AI models, but that's to be expected, given that you don't pay for it.
It has at least improved with the second category recently that it seems to be more right than wrong. But it still tends to work your query into whatever results it got, which makes it downright dangerous for when people are doing searches on health-related information.
Taking a response at face value is just asking for trouble; they're useful primarily for finding references. The summaries are a good way of prioritizing what to read, and if nothing else the documents it cites tend to be relevant.
The people who get burned by these things are the ones who just blindly trust the generated text.
Also some of the skill is learning to feel out when you can trust it and when you can't, but if you're just starting out, always assume the bullshit machine is generating bullshit.
I still find Claude better for learning math but I won’t use it for other everyday things again.
As long as there is money to be made by being the first thing pushed to a consumer's eyes when they want information, search is an adversarial market, doomed to cat and mouse games and failure states.
The early days of PageRank were entirely anomalous because nobody was poisoning the well they were using yet. The second they made doing so financially incentivized, that situation went away and Pagerank and all algorithms like it were doomed to struggle.