[delayed]

Very rare for me to use Google nowadays for something niche. For those cases, I would assume it requires multiple sources to get a definitive answer, so it goes to ChatGPT.

[delayed]

Why do you think there’s a high error rate for general queries nowadays?

[delayed]

Yeah fair. I guess I have a very different experience, as I mentioned above, I basically never use it for any niche/hard queries that depend on multiple sources or research. For all my common queries, it's reliable.

[delayed]

Sorry, not going to take the bait here. I would argue in 2024/2025 it was dogshit, but it really isn’t nowadays. You can even see the sources for most of the summary queries, like literally on the same line as the bullet point.

My google queries are like different combos of weather, specific event questions, directly attributed to wikipedia data and etc. LLMs are way past the point of reliability when it comes to data extraction from static websites.

> LLMs are way past the point of reliability when it comes to data extraction from static websites.

Two very modern counterexamples:

https://app.opencve.io/cve/CVE-2026-83844

https://www.sentinelone.com/vulnerability-database/cve-2026-...

The information they feed from is simple, clear, and static. And yet the both of them are hallucinating mitigations, compromise indicators, and other details of the vector that flatly don't exist.

The question GP posed wasn't bait. It's still a blatant problem in 2026. It's not just shitty cheap models, either. Just this morning I caught 6.1 Sol (via Codex) incorrectly answering a question about the codebase it was actively working in.

That is fair. I always ask for sources whenever it is something I am unfamiliar with instead of just trusting the response, that way I can go and double check and it ensures the agent at least read something on the topic

I've found the value of this approach has diminished over time - now I need to go and read the source and try to identify if it was generated by an LLM itself.

> And for any meaningful query, it’s straight to ChatGPT/Claude.

What a disgusting use of technology. Kicking a query to an agent, that's leveraging a model that has bulk of humanities information in it, so it can use a web browser to go looking for the thing you searched. WTF.

As opposed to Googling, which will

1. leverage the AI Overview model (with the IQ of a brain-damaged monkey),

2. run an ad auction and populate the first 6-10 items from that,

3. then do a perfunctory exercise of searching its spam-riddled index to round out a few results at the very bottom of the page.

Yeah, that guy asking ChatGPT is a monster.

You forgot to mention all the well-documented ad auction fraud that came out during the anti-trust trial (they defraud advertisers and sites and do some shady stuff to users —- on every impression).

Yeah! Reading stuff written by people sucks!

I never said that? But when i’n searching for a certain answer, I don’t really care about paragraphs, posts, etc. A good chunk of time it might be data auto-generated from big databases too.

Time and place for every tech/use case. When I’m in the mood for reading long pieces, i do that deliberately.

I actually like to discover some lost articles that I vividly remember and it's honestly great for that.

Just use a decent search engine dummy.

What a dumb reply

Dumber than I'm so stupid I'd rather read llm generated garbage that massages my ego... What an insightful question, you're really thinking about this the right way!

Why am I not surprised by that reply.

Because you are an idiot

We've banned this account for repeatedly breaking the site guidelines and ignoring our requests to stop.

If you don't want to be banned, you're welcome to email [email protected] and give us reason to believe that you'll follow the rules in the future. They're here: https://news.ycombinator.com/newsguidelines.html.

We've banned this account for repeatedly breaking the site guidelines and ignoring our requests to stop.

If you don't want to be banned, you're welcome to email [email protected] and give us reason to believe that you'll follow the rules in the future. They're here: https://news.ycombinator.com/newsguidelines.html.

We detached this subthread from https://news.ycombinator.com/item?id=50385425.

Except the summary doesn't fucking work. It's consistently wrong.

As I mentioned above, if you’re making a niche query, maybe. If you’re making a very generic query “average temperature in Hong Kong in January”, it’s fine. You’re overestimating average complexity of a query.

"Fine" may be "right 95% of the time" but I'll look like a jackass 5% of the time that I trust the Google AI Overview.

And I won't know which ones are the wrong ones, so I may as well trust it 0% of the time.

Most people I know use AI Overview now as their primary way of doing any sort of research or question-answering... and I urge them: If getting an LLM take is suitable for your purpose, at least ask free-tier ChatGPT, which is at least an order of magnitude less likely to vomit up false or confused information. Also, asking actual Gemini is another sane option that is far superior to AI Overview.

Does it really matter that much? Did it matter that everyone's parents pre-internet were wrong half of the time?

Most people are not asking google to build them a train signaling system but to look up some question on Wikipedia for which the answer does not matter that much

If that’s the case, those people should just install a 7M param model on their phones and use that. It’s probably as accurate, but works offline.

Sure but that is harder than pressing a button or saying hey Google. It requires you to think, make a UI, etc.

Think of the incentives and the path of absolute least resistance/thinking/energy use and you will arrive at the product people want.

I think you can press a button + say "Hey Siri" at this point (though that assumes Siri works at all, which I agree is a stretch.)

7M param doesn’t do web searches for you. Otherwise, yeah?

7M is fine if you don't care about the quality/correctness of the output and just want some utterance that looks plausibly related to the question. I guess that is the bar this thread has set for Google search quality being good enough for most users.

> Does it really matter that much?

Does truth matter in the post-truth world where disclaiming truthful news about you is as easy as calling it fake?

Interesting, I had terrible experience with ChatGPT as search not-so while back, found gemini quite useful for finding out the articles and brainstorming information.

> but now Gemini summary pretty much covers all of my use cases. And for any meaningful query, it’s straight to ChatGPT/Claude.

How do you live like this?

I use it much less than I did prior to LLM becoming useful. However, when I need to search, it just works. Their AI integration with their search is also great, but I couldn't justify it and ChatGPT subscription. I can't bring myself to cancel Kagi entirely, though. I still did 600 searches this month.

It's a AI answer and a huge page of ads

For me, it took 30-60 seconds to notice dark mode preferences on my browser, but popped up an ad saying that search is garbage without the app, and that I should install their malware on my phone.

When I dismissed the popup, it produced much better results than screenshots I’ve seen from logged in browsers.

I guess that’s an improvement?

Yeah, agreed. If Claude/ChatGPT would load and respond faster, i’d go for them. But Google’s speed is unmatched for quick queries.

Every time I get a Gemini summary it reminds me not to use google. It's almost always wrong about key data.