Hacker News Reader

Comment by redsymbol | Hacker News Reader

The most interesting part of the official announcement [0]:

"As people gravitate toward firsthand perspectives..."

That implies Google is seeing a shift away from people relying on just AI-generated summaries, and checking the cited sources more.

I have certainly found that when I ask a chat tool something, and then go to the human-written source webpage, it's not uncommon I get a different insight than what the AI implied. Not always of course, but often enough that I have noticed it.

[0] https://developers.google.com/search/blog/2026/07/search-con...

View on HN

> "a different insight"

What a generous euphemism. Depending on the topic, it's not unusual to see Google AI search referencing links that indicate the exact opposite of what it presents in its prose.

So many of these tools are presented as if they're doing something you'd intuitively expect of a human operator, like collating search results and then summarizing them in this case, but the actual operation of them is so alien to us that our intuitions don't apply and these presentations are all but fraudulent.

Because it's presentation cites references, and because those references sometimes reinforce the expression of the AI prose, we grant it trust, but it doesn't use those references in the way a human would and so the relevance of them being cited is radically different (and weaker) than we're accustomed to.

The closest example in human behavior might be the rushed, naive student who just pastes citations into an already written paper at the last minute, but even that's an anthropomorphism that misrepresents the alien ways in which the tool works.

Yesterday ChatGPT gave me 5 references to a site that doesn’t exist. It’s such a bummer how useless these tools are so much of the time, while seeming so helpful. In fact they are actually helpful, which is why it’s so irritating.

Meanwhile, I asked ChatGPT to verify the ALIAS DNS records of its search results so I could build evidence Netlify was blocking me [1], and it worked flawlessly.

I've replaced Google almost wholesale with ChatGPT and Claude.

[1] If you use Claude Code to assess a Netlify website (even your own), they'll block your IP address from the entire netlify.app IP range.

Hope they move away from the engagement model. I prefer blogs over reddit/insta/tiktok in search results.

I don't think that means anything. It's just some random words to add pretense to an arbitrary change.

I mean, just stop to think about it. You have a Youtube account, you post videos. Analytics shows traffic from Google but can't show you the keywords.

At one point, Google merged Youtube with Google+. Just did it outright without asking anybody.

Youtubers link their Youtube accounts with Google AdSense.

You can link your Youtube account with your Google Ads account if you want to advertise.

One of the latest features added to the Search Console was... an AI chatbot.

So Google showed us they can link anything with Youtube when it wants, and they can add new features to GSC if it's AI, but for some reason showing Google search terms for Youtube channels took 20 years.

I wish I knew a word like enshittification but for when something improves but at a rate so slow you don't even feel like celebrating it anymore. I think I'd use such word a lot...

The term you're looking for might be "maintenance mode". This sounds like what we do with software we don't care to develop further, even if it's the big cash cow

That doesn't sound bad enough. It feels closer to "abandonment mode."

Hopefully that is the case, because what Google and others are trying via AI, is to privatize the web. They already tried it via AMP.

People are probably annoyed that AI/LLMs are lying to them, so not everyone sees the main threat being Google trying to containerize and "walled garden" approach access to information.

> I have certainly found that when I ask a chat tool something, and then go to the human-written source webpage, it's not uncommon I get a different insight than what the AI implied.

Well, many people pointed out that AI lies. We can have two reasons for this:

1) deliberate 2) a feature of how AI works

I think it is mostly 1), but 2) could also factor in, because if ALL AI is lying, then probably something is wrong with how AI works right now. Since they prefer to hallucinate. But AI could be made stricter and not hallucinate, so I think 1) is the main issue.

> But AI could be made stricter and not hallucinate

Do you have a source for this? So far every single LLM-like does a lot of hallucinating, often just small things like wording that implies stronger correlations than exist, or wording that implies relationships that don't exist, etc. but sometimes it's also just wrong facts.

Which lab is going to stop AI from hallucinating first?

What’s taking them so long? If OpenAI was going to do it, why didn’t Anthropic do it first, or vice versa? Or if we assume collusion: shouldn’t DeepSeek have released a hallucination-free LLM to steal the US’s thunder?

(Good points on AMP and walled garden)

the whole point of an llm is to generate "plausible sounding" tokens one after each other.

It can sadly not hold any concept of truth or reality as we humans understand it

We’ll be awaiting a source a while! https://news.ycombinator.com/item?id=48873333

Reply on news.ycombinator.com