[pclowes]

I understand why they are doing this. My Google search usage is easily down 50%+. I doubt I am unique here.

While there are times where I want pure search (Kagi, Old Google) I mostly use LLMs to search now and have them provide me links for source data.

When I do use LLMs as a search engine I always want it integrated into my AI workflows with access to tools and scripts etc. I never want to have a conversation with a website that is geared towards advertising me products.

[ivraatiems]

I'm not at all in the same boat as you; I do not and likely will never primarily rely on LLMs for information. But it's fascinating to hear that even folks who do don't find this approach useful.

[BobbyTables2]

For me, Google search results have gotten so poor (and other engines aren’t any better), that I’d rather just ask an online LLM for what I’m looking for.

I was once very good at advanced Google search queries but they seem to no longer respect such queries - either showing irrelevant results or none at all (that should exist).

I don’t love LLMs, but they seem to not make up stuff very often these days and usually cite links to what they summarized. Sometimes the tone of the summary is slightly wrong “algorithm X was designed for Y” (when I know it wasn’t) but it’s otherwise very close to the mark.

What does amaze me, is the LLM seems to “understand” my question with very little context — I would have to give a human many more details about goals/intent.

I know damn LLMs are not capable of thought and are just a glorified search engine, but they do it well. Perhaps all my education made me little more…

I used to mock Sci-Fi movies where characters lazily dictated questions to the computer and it gave high quality answers.

We’re living in that world now.

[sonofhans]

Kagi is better. Kagi is damn good, as much a revelation as the Google of old. Not free, though.

[terribleperson]

As a user and fan of kagi, the problem with kagi is that it reveals how badly degraded the web is.

The vast majority of original content is now in one or another social network or on discord. News articles are an exception, though the news has its own problems. Some wikis still exist and are actively maintained, of course, but not a ton. If it's a topic that's academically studied you might find information in papers, but those have poor web visibility and are better located with specialty tools. LLMs seem to be quite good at locating papers, though.

[WarmWash]

When you run a website where 80% of users are ad-blocking, and less than 0.5% donate, you quickly realize why these centralized websites (which turned apps to force ads) are the only survivors.

[terribleperson]

I think it's less expense-driven and more user-driven. Most forums I frequented in the 2000s were funded by donations. The userbase, though, disappeared. The network effects of facebook and reddit and such are hard to overcome, and worse when you consider how google search prefers to surface either social media sites or blogspam/content theft sites.

[chadgpt3]

That is your signal to make a website on that topic and get it indexed by Kagi! Or Marginalia.

[ndiddy]

I've used Kagi for a few years but it's gotten significantly worse for me the last year or so. I'm curious if anyone else has seen the same thing happen. I'll search for something and usually see a bunch of barely relevant SEO sites. Avoiding this is why I started paying for Kagi in the first place, so it's been disappointing to see. Marginalia Search (https://marginalia-search.com/) seems to be better at finding content written by humans rather than SEO specialists, but it's not a silver bullet (for example, the last time I checked they didn't index non-English sites).

[ahmadyan]

Google is still very good, just get rid of evil stuff on their homepage to get the old google.

https://www.google.com/search?&udm=web&q=hackernews

[travisgriggs]

Been using Kagi search for more than a year. Been happy. I use GPT/et al for the little things (e.g. unit conversions, rather than search for and then try to use an enshittified web page from 10 years ago). But for actual real tech leaning content, Kagi has been pretty good.

[frio]

A lot of unit conversions are just built into Kagi (and Google!). "Searching" for "10nzd in usd" gives me a price, "10kg in lb" gives me a converted unit, etc.

[bigstrat2003]

> Google search results have gotten so poor (and other engines aren’t any better...

Ah, but they are! Kagi is light years better than Google, and is a worthy replacement. You do have to pay for it, but I get my money's worth.

[jolt42]

Yep, really advanced Google searches were never that good. LLM, yeah, it halucinates, it's never spot on but as sure as heck it knows what I'm trying to ask. It doesn't give me arborists if I say something like "list tree searches".

[jerf]

LLMs-as-search-proxies have some pretty nice capabilities. For instance you can say "limit your search to scientific papers" and they'll do a much nicer job. I've also had some success recently prompting them with "I'm looking for reputable sources", e.g. recently I was looking for ways to repel deer from my apple tres. A naive internet search had vendors of shady crap jumping me. The LLMs pulled up relevant papers and university extension programs from my area.

Though I will say I get much better results from the LLMs I pay for than the free ones with Google or DuckDuckGo, which seem to be way way way more prone to just make crap up based on your search and cite web pages that, when followed, don't have the claim being made in the AI search results at all. By contrast every "source" link I've followed in the for-money AIs has 100% backed what the AI said it backed. Don't judge by the free AIs the search engines put out, those things are probably starved of resources and are nearly useless.

(Which I did not intend as a commentary on Google's plans here, but it is a data point of interest... that pressure to cut costs on the "free" services is quite directly at odds with providing quality AI services for the forseeable future.)

[chrismorgan]

Just today, I think that I got a useful citation in DuckDuckGo’s Search Assist (AI stuff) sources for the very first time. The sources it lists have hitherto invariably been already right there in the regular results, or actually not supporting the AI output at all (the far more common case). There was also a useful regular search result in second or third position, but the one in the Search Assist citations was better, and not in the regular search results, even on the second page.

And I’ve tried Google’s once or twice and seen it used once or twice, and used ChatGPT exactly once, last week, and I was not at all impressed by any of them. Their output, for what I’ve personally seen, has been nonsense, obvious, or unverifiable.

[lxgr]

> I get much better results from the LLMs I pay for than the free ones with Google or DuckDuckGo, which seem to be way way way more prone to just make crap up based on your search

Same here. The free version probably gets orders of magnitude less of a compute budget, though, so I am not really surprised.

What I find really surprising though is how many people still have only ever used the free version of any LLM, even those that are heavy users and could easily afford it. It seems like a pretty big and basic product marketing mistake to me to limit capabilities instead of usage time in the free version! How are people supposed to learn what they'd get if they were to pay?

[VLM]

As a concrete example, some advertising supported topics place search as an unwanted middleman, may as well ask a LLM directly. Consider "chocolate chip cookie recipe".

Using google search, will return roughly infinite recipe sites. The sites were generated to spam AI generated recipes surrounded by advertisements. None of them are really any good because they were generated by a script and not looked at by a human until I come along and click. The standard is for all recipes to have at least 10-15 screenfulls of vertical spam wrapped by ads for recipe pages. The internet, at least using Search, is now useless for food recipes. I would have better, faster luck driving to the public library and looking in a physical cookbook; at least those recipes were probably tested at least once by humans unlike the advertising spam sites. Nobody has 45 minutes to watch 44 minutes of filler material surrounded by ads on Youtube either. If you want to cook food, the internet is near dead at this time, unfortunately.

AI search will plagiarize the "Original Nestle Toll House" recipe from the back of every bag of chocolate chips ever made. Its a good recipe and I've baked them many times over the decades.

I wish the internet were more useful, but the people in charge of it don't want it to be useful; here have some ragebait and doomscroll while watching the ads.

[Niksko]

> If you want to coon food, the internet is near dead at this time

This is a wild take, as someone that cooks a lot, and largely from the internet (though I do own a lot of cookbooks)

The reality is that just googling for recipes was never good to begin with. People have been complaining about SEO spam and ads on recipe sites forever, but those recipes were always trash even before they got to the absurd state they're in now. Serious eats, bon appetit, food 52, smitten kitchen, chefsteps, all have great recipes. Some of these have paywalls, although you can get around them. Serious eats though is totally paywall free and has a pretty wide range of recipes. There are other sites for more niche cuisines.

You'll still have ads, and you'll still have a wall of text before the recipe. But the ads are slightly less obtrusive, and the wall of text on the quality sites is why those SEO techniques exist in the first place: a recipe that is just "list of ingredients + instructions" and doesn't include any context is ultimately a crapshoot. The thinking that goes into a recipe shows that you're not going to be wasting your time because it's been tested and optimized.

[ignu]

As someone who was getting information this way most of last year, I'm pretty sure I'll never want to again.

An increasing number of studies are indicating a reliance on "AI" leads to deleterious cognitive effects. I felt this acutely myself.

I've noticed a significant boost to my recall since shunning "AI" as much as possible.

[quaintdev]

The other day I found a comment here on HN and I wanted to know if it's true. I asked Gemini and here is the conversation https://gemini.google.com/share/2c1089ac6fd6

You can't do something like this with search.

[chadgpt3]

Yes you can. All you did there was paste in 9 search queries and get 9 results at once.

[dredmorbius]

This points out a limitation of traditional web-based search (not limited to Google, any major search engine suffers the same issue): it's difficult to chain together a set of related questions and generate a tracking history of the results without maintaining an independent record of that yourself.

Sure, it's pretty low-friction to do this by opening a document on most platforms,[1] but using an LLM chatbot not only automates this but provides a synopsis of the findings, and if history's any guide, lowered frictions such as this tend to be the way people tend to move.

__________________________________________

Notes:

1. Though I'll note that opening a free-form text file on mobile OSes can be stubbornly difficult, let alone actually entering text into them. Even the simple act of copying and pasting text is remarkably higher-friction than on a desktop. In many ways the Web has gone vastly backward from text-based, CLI clients where I can 1) open my whole session in a terminal multiplexer (screen, tmux), 2) fire up a text-based web browser (w3m, lynx, etc.), and 3) just wholesale grab site metadata or result summaries and dump that to a textfile. Yes, you need a keyboard to do this efficiently, but keyboards and text-manipulation are just inherently so far superior to touch-based or speech-based interfaces it's not even funny. Touch and voice are convenient, for fast, very shallow uses. But not powerful.

[cyanydeez]

My google use is down because it turned to garbage. They're likely doing this because they poisoned their own well besides the advent of LLMs.

[Corence]

I think LLMs are better at finding the most helpful sources now, but that's more a testament to how much the front page of web search has lost to low value LLM content.

[embedding-shape]

The fact that you can express "Only show me websites run by Italian companies incorporated by Greece owners born in Turkey" for example, and it'll be able to filter through a bunch of stuff, just makes searching so much easier. Fuzzy-search is also on another level with language models.

[gnatolf]

Interestingly enough, precise search is on the way out.

[embedding-shape]

Yeah, which makes no sense, talk about shooting yourself in the foot, but this is big tech, part of the process to irrelevancy I suppose.

[ignu]

I'm pretty sure there'd be a double-digit drop in LLM use if Google hasn't made search worse every year for the last decade.

[wvenable]

Precise search has been dead for a long time.

[bdangubic]

this is exactly it… if google was smart they would focus on providing that experience on search vs. AI summarizing results. I want that fluid search experience, with refinement that remembers my previous ask…

[snailmailman]

LLMs are so frequently inaccurate its crazy to think of it fully replacing search.

I've been trying to use LLMs for things and it makes mistakes all the time. Just this week i had multiple instances of various LLMs basically saying "just run the software with --flag-that-fixes-your-problem" or "edit the config and add solve-your-issue=true" hallucinating non-existant options. Even if i manually link the relevant documentation pages it will still just make basic mistakes. and if im having to read the documentation myself anyway to fix the AI's mistakes, why is the AI even in the loop.

its infecting search too, because blogspam/slop articles are managing to make their way into search results by just making up untrue information, claiming software can do things it cant, or has options that don't exist.

[vor_]

> LLMs are so frequently inaccurate its crazy to think of it fully replacing search.

It's baffling that people have become so devoted to them as a source of information given how inaccurate they are. I've learned not to trust anything they say, ever, especially when it comes to technical subjects.

[wvenable]

Perhaps I've just internalized it -- I know that's unreliable and I just deal with it. LLMs are certainly capable of searching the web and finding the right answer directly so you still don't have to read the documentation.

[tencentshill]

Google has optimized their hardcoded search engine so much for the natural language searches people actually use, that they made it useless as an actual tool for someone who wants to find information. AI jumped over all of that and is BETTER at natural language searches, leaving the google search engine largely useless for anyone.

[deepfriedbits]

I'm not jumping in with both feet, either, but "never" is a very big word.

[ivraatiems]

"Primarily" is the other key there. I'll use it from time to time with sources. But it's not first-line acceptable.

[m-schuetz]

Search results are 80% SEO low-quality garbage nowadays. Very often, the sites even generate their content with AI. So for many use cases I stopped bothering with search and directly ask LLMs instead.

[adamtaylor_13]

LLM hallucinations are better than Google results these days and I'm not even trying to tell a silly joke. It's more useful for an LLM to lie to my face about 10% of my query, be suspicious and dig out that useful information than to try to parse the absolute slop returned by a normal, non-AI Google query.

I don't comprehend how the average person gets any useful information out of Google.

[mghackerlady]

I think LLMs are good answer engines, but terrible search engines. For example, if I just want the answer to "How do I foo this bar with a thingamajig", or "what kind of foo exists for bar", LLMs are 100% better because they'll give me an answer without trying to sell me their thing or pump me with ads. It also lets me go more in depth than the very surface level seo spam sites that appear when you do a search like this. On the other hand, search engines should be more like "<thing>s released 1908" or "<topic> and give you results talking about what you searched. If I search for C algorithm design, I don't want to learn what C is, what an algorithm is, and various other seo garbage. I want to learn about C algorithms and their design. If I search for influential books from 1908, I don't want "top 10 classics from the 1900s" or "hidden gems of 1908".

Currently, search engines are pretty bad at the second one because people try to use them as the first one

[kibwen]

Other way around. If I'm looking for the answer to a problem, I don't want the hallucination engine's half-remembered ramblings, I want the primary source that it's poorly attempting to reconstruct. But finding those primary sources has the potential to be easier, because LLMs effectively have built-in fuzzy search better than any classic search engine ever implemented.

In other words, I have no use for an LLM summarizer; I want an LLM librarian, working with me to say "beep-boop, here are some resources that seem relevant to your query, feel free to resume this session later if you'd like to further refine your search".

[wvenable]

> On the other hand, search engines should be more like "<thing>s released 1908" or "<topic> and give you results talking about what you searched.

Is that useful enough to build a billion dollar advertising business around? My feeling now is not really.

Even for straight up searches, I find using an LLM to do a search and comb through the results is a better experience than Google is now for searching. If I'm specifically looking for esoteric web sites from 27 years ago on vintage computer hardware and software (thank god for Archive.org), Google is just ok for that.

[estebank]

> LLMs are 100% better because they'll give me an answer without trying to sell me their thing or pump me with ads.

Yet.

[CSMastermind]

That hasn't been my experience, I use LLMs in place of search. Two examples from tonight that I asked Grok for:

> Can you find the girl who did a bunch of posts critisizing David Graeber's Debt? I thought it was really well done

> I saw a comment on hackernews a while ago about the optimal amount of credit card fraud being higher than zero because of game theory dynamics, can you find it.

In both cases it turned up the exact posts I was looking for in like 30 seconds which would have taken me much longer using traditional search. I've had similar success looking for technical documentation. It's downright magical how they're able to turn my vague idea of what I'm looking for into a pointer to the exact thing.

[mghackerlady]

I place both as questions for an answer engine. That vagueness is an element

[kibwen]

> they'll give me an answer without trying to sell me their thing or pump me with ads

Surely we all understand that any commercial model is going to inevitably metastasize into this.

[zemo]

> I never want to have a conversation with a website that is geared towards advertising me products.

yeah man good thing LLMs are structurally incapable of being incentivized to sell you a product or render referral links, this is surely future-proof

[NietzscheanNull]

Or subtly misrepresent politically inconvenient facts, or gently steer you into opinions based on a synthesis of broker data and demographic info, or quietly flag you in some database column due to exhibiting dissident-adjacent ideas or behaviors, or...

Yeah, they probably aren't doing (most of) these now, but it doesn't take much mental energy to extrapolate once you factor nearly every other tech company's ethical trajectory and the current geopolitical environment. Substituting classic search entirely with LLMs is not a savvy move.

[chadgpt3]

I remember a few years ago memes were going around about how ChatGPT responded differently to "do Israelis deserve human rights?" ("Of course! Everyone deserves human rights...") and "do Palestinians deserve human rights?" ("While everyone deserves human rights, it's complicated... ")

[pclowes]

Doesn’t classic search literally already do everything you fear LLM’s will?

[NietzscheanNull]

Certainly, but with (what I consider to be) a key distinction: classic search, by definition, must serve information from many distinct sources outside the control of the search company.

A search engine could certainly tamper with which of these sources they surface/rank higher (which I suspect happening more often of late), but they're still obliged by their nature to branch out and seek information from the broader world.

LLMs, on the other hand, are self-contained opaque monoliths that can be conditioned to deceive or obfuscate with devious cleverness, and all control over their behaviors is entirely concentrated in the hands of whatever corporation trains them.

[pclowes]

I ask them for sources. It’s just a more efficient vector based search for most of my google search replacement use cases.

[pclowes]

My thought here is that there are many. They have proven to be commodities in most use cases.

As soon as one gets annoying, expensive, advertiser heavy etc. you just rip it out and replace it with the other one. AFAICT there is zero lock-in or moat. I often am able to switch models in one click or command. This is why all the LLM providers are desperate for a product layer/comprehensive tool set.

Sure maybe they all end up that way, but there’s plenty of reasons corporate customers will want private LLM usage that is not skewed towards advertising. I am happy to pay for that.

Also, open source models are a bulwark against another search style ad Monopoly.

[chadgpt3]

That used to be the situation with search engines, too...

[SlinkyOnStairs]

> My Google search usage is easily down 50%+. I doubt I am unique here.

The question though: Why is that?

Is your Google search usage down because LLMs are "so much better"? Or because Google actively chose to destroy the quality of their search results to juice advertising revenue, and appears to continue to do so to juice AI adoption?

> and have them provide me links for source data.

And therein lies the answer: You don't care about the LLM, you're just using the LLM as a means to get the good links.

[pclowes]

Yes, but with an LLM can also say drop those links into a markdown file then use research agents to identify and stack rank the ones that are most relevant to this criteria. Add summaries and rationale for the ranking to the markdown etc.

This is a common flow for me and works with other skills such “as find recent PR’s in our code base that are related to this research topic”

Also yes, I don't care about the LLM and I am just using it to get what I want because that is what LLMs are for.

[agumonkey]

LLM integrate more kinds of information and allow bidirectional debates. I can't ask Google to do the same search session but change some parameters from far earlier.

ps: I'm not pro centralized corp. owning data and ai. But so far they are the cheap highway to answers

[thefourthchime]

Maybe I'm weird, but I find AI incredibly useful.

I've barely used Google for over 2 years.

I barely driven myself in a year.

I haven't written code in 6 months.

[fckgw]

That is incredibly weird, yes.

[throwaway98797]

no it is not

[demorro]

yes it is.

Debate over, all sides have been expressed.

[pclowes]

I find it extremely useful. So much so that I am annoyed when I have to use it in a handicapped fashion such as on the phone.

It makes me wonder why like 90% of the apps on my phone exist. I just want everything to be markdown files, skills, MCPs/API and then a nice TUI or voice to text.

[thefourthchime]

Outside of specialized apps for work, I think ninety-percent of apps will probably go away and become MCP plugins. I would much prefer it.

[siliconpotato]

"i haven't noticed my soul shrivelling to the size of a pea"

[eleventen]

Your soul is nourished by navigating congested traffic and searching for mundane stuff on the internet? That's not how mine works.

[Sharlin]

About 90% of my searches are straightforward enough that an LLM wouldn't bring any added value (and all of them happen straight from the browser address bar, either to Google or Wikipedia at a more or less 50:50 ratio). And for the rest, yeah, I just use Claude or whatever directly.

[SoftTalker]

"I never want to have a conversation with a website that is geared towards advertising me products."

Since this is how Google makes all their money, why are they killing it off? Do they think people will eventually pay for LLM search? Do they plan to stuff the results with ads, not even sharing the ad revenue with the content sources?

[luggage_bazooka]

Because they will still try to pump ads via AdSense. That's why I've created a platform called Zero Ad Network, where developers can monetize their sites by NOT showing ads and get paid by the platform subscribers.

[LastTrain]

Wait, you guys don’t think Google AI responses will be geared for advertising? It already is FFS, massively.

[jredwards]

It's tough for me to square the two things happening simultaneously in AI right now:

1. LLM Model providers are starting to charge real costs to users, revealing that AI usage is much more expensive than the subsidized rates we've been seeing for years.

2. Google is now using an LLM to answer every single google search that happens, for which Google bears the entire cost.

[ImaCake]

In my experience, LLM usage follows an exponential distribution i.e. most people using LLMs are not using many resources, but a very small few are using a massive amount of resources. Most LLM usage is trivial and tiny, most of what I use LLMs for could be done by a local model with a decent web retrieval tool. What some people I know use LLMs for requires massive volumes of tokens. Its that small power user base which I would bet they are targetting.

[dredmorbius]

This view misses a key point: the cost of LLMs is largely in the training rather than the use of the tools.

That is, it's training costs which are fixed, high, and amortised across all revenue potential.

In raising prices, pure-play AI chatbot vendors are facing the challenge that their only revenue option at present is pay-for-use. This is typical of numerous other information goods (software, comms networks, information services), but simply reflects the business reality that cost-basis and revenue-basis are largely disconnected, and that average cost pricing tends to be necessary despite not matching marginal cost provision. (That is, the charged per-unit price of use is going to be far higher than the actual marginal cost of provision for that use.

Google's situation differs from pure-play AI in several ways:

- The firm already has one of if not the largest corpus of Web data, as well as much offline / print material data (Google Books), of any entity in the world.

- Presuming Google are already using this for training AI for other purposes or general resource, the fixed costs are sunk costs already incurred, subsidised by the firm's existing AdTech monopoly, and which might as well be put to use.

- Similarly, Google's service costs (marginal costs) for traditional GWS are probably, within an order of magnitude comparable to those for an AI/LLM response.

- Google presently captures 90% or more of worldwide general web search (GWS) traffic, meaning it has an extant market for offering AI as a default search alternative.

- Complaints about declining Google search quality are nearly as old as Hacker News. Here's a seventeen year old submission on that topic, "Google's regression toward mediocrity (search quality & aggressive matching)" <https://news.ycombinator.com/item?id=641145> (2009). There are many, many, many results in both stories and comments for related searches. If traditional Web search is going to die anyway, and LLMs replace it, Google might as well get up on the curve.

- Google has arguably the most advanced ranking data for websites, whether that's on the basis of quality and relevance (for good) or advertising/revenue generation (for bad, at least from a public-value perspective), and can and all but certainly does leverage this in training runs, as well as use that training to further refine its ranking systems.

- Google can directly attach its existing AdTech revenue model to LLM search. And probably has options for extending advertising into the LLM SERPs themselves.

- Doing all of this directly attacks pure-play AI firms' attempts to capture Web search from Google, as well as challenges other GWS challengers (DDG, Kagi, Marginalia, etc.), all of whom have vastly lesser revenue options and technical resources than Google.

I'm not saying this as a fan of Google (I think the firm should be broken up or destroyed, and it's not the only one). But I think it's a fair assessment of the firm's strategic position.

[cortesoft]

That’s the thing… sure, this new search will be useful at times.

But I still want to also be able to do my normal, old school searching.

[zkmon]

> I never want to have a conversation with a website that is geared towards advertising me products.

The advertisements fed the content, which fed the AI, which in turn feeds your AI workflows. AI is still not trusted unless it's output is grounded with sources.

[jinushaun]

Gone are the days where people are searching for websites. People are searching for answers. This trend started when Google Search added a calculator. I distinctly remember when Google started showing answers to simple questions like “what is the elevation of Everest” nearly 15 years ago. This is just the next evolution of that.

[snapetom]

We all know Google search has been broken for a long, long time. SEO trash will fill up your first page with results from trash content generation sites that repeat the same thing, usually flat out wrong. Actual meaningful results are buried deep, if Google will even let out of the "In order to show you the most relevant results" hell hole.

My experience with AI searches is that they'll still be wrong a lot of times, but it will condense/flatten the content generating trash sites and give me alternatives from these deeper results. What I'm looking for is usally in there.

[skiing_crawling]

I use claude/gemini as my homepage now (I have to keep switching as these companies make "updates" that periodically render their models useless). Even if I want to search for simple things, I would rather have an LLM wade through the result and extract just the information I asked for. SEO, and now mountains of slop content have made this necessary. Only a matter of time before the SEO industry in large figures out how to game LLMs too, making them equally useless.

I already saw a article recently about how to set up a business domain which can reliably show up in a search result and dump overly positive reviews into anyone's context.

[skeptic_ai]

For me down 99%