Why Google search is falling apart [video]
youtube.com
youtube.com
If friends complain about that and how growth-first culture ruins good products, I buy them their first month--Kagi seems to epitomise finding a niche and serving it excellently. If you're technologically-savvy, don't like your time wasted and are willing (and able) to pay for search, you should try it out. (It's not surprising that they're a bit of a meme here on HN.)
[1] https://help.kagi.com/kagi/why-kagi/why-pay-for-search.html
And many people, as evidenced by Google’s profitable enshittification, do trust the Google results. To them, if they got scammed by an SEO-hacking site, or a site paying Google for search result placements, it’s just how things are on the internet. They would not understand the complicity of their search engine.
All together, that’s the magical thinking Kagi will have to compete against. Because early adopters come and go. They will need to transition to mass market adoption, at least at some single-digit percent of the market share.
Disagree. There is plenty of money to be made in being the premium search engine to the rich and educated. To the degree they might solidify that font, it’s in pursuing academia.
Edit: Reading the sibling comments I pay for the $10/mo unlimited and use it liberally for about 2k searches a month.
- Maybe the localization of search results doesn’t work well in Europe,
- Sometimes it’s the literal name of the website and Google is better at that.
I'm a Kagi user of 2 years and it's actually not the default on my phone. If I'm on my phone, I'm likely using it to search local businesses and nothing can compete with Google's business + review data.
I've been using Kagi quite successfully to recommend restaurants in an area. (Quick Answer.) Sometimes it's just summarising a Tripadvisor. But often it's pulling in recommendations from Reddit and personal blogs, which greatly improves the reliability.
Can definitely see them having issues in Europe, though, particularly when it comes to crossing languages.
Honestly, the fact they're so unwilling to drop the axe on large sites scamming them probably had the biggest effect here. If the likes of Forbes saw their whole domain get blacklisted, we'd see a lot less scammy bullshit end up on page 1. But Google don't want to do that, for various reasons.
Let's say you had a Mission to "Organize the world's information, and make it universally accessible and useful".
Let's also say you had a Prime Directive of "Don't Be Evil".
To help implement those, for anyone who won't get aligned, let's say you operate a colorful company shuttle bus, which can remove them from campus, in a one-way trip.
Wouldn't this solve a whole lot of problems, and narrow the solution space for a lot of other problems?
Now that even a tiny additional fraction of that information, along with the obvious desire to manipulate and control some of it, has arrived this model completely falls apart.
It was doomed from the start, and there was never any connection between organization information, and actually solving real world problems. No surprise the world is in worse shape than it was before and Google has a bunch of money it has no idea how to actually spend on the core problem.
They just don't have any massive advantage to do so.
Their goal has ways always been capture. Keep us on their portal. Well they're succeeding, because they fill pages of response with their own data, and the alternative is to click poor results.
They have a vested interest to keep you on their page.
It doesn't matter that their buffonerific answers are often gibberish, and so many times very wrong. Or worse subtly wrong.
Those answers keep 99.9% of the people on Google only.
And just take another look, too. How many of the responses aren't just their horrid AI junk, but many of the rest are youtube.
Why spend time fixing main search, when it being broken pushes people to stay on google.com, or go to youtube?
They can fix it, they just don't want to.
It’s only worth it if they’re confident in their own ChatGPT.
There is simply _too much_ information. You cannot economically approach this problem. It doesn't matter how clever your staff is.
What you're describing is the fact that the _stated objective_ and the _profitable outcomes_ are two different things. We've all known that since they originally printed it.
The issue is the inclusion and ranking of easily discernable junk. Google has stopped caring, for my cited reasons, and can tackle this if they want.
There is nothing preventing this other than it actually helping them by making their products preferable.
For example, eliminating any page with referral links. Eliminating copycats. No, it's not an issue for them. This alone would get rid of 1/2 the junk.
But that would reduce their silo. And further?
It would mean that other search engines which buy Google search results would benefit too!
Far better to show their picks on their first page results, whist keeping lower quality answers in their general search results. They can show competition exists, without really helping that competition, and people stay siloed.
The flaw itself is in the standard business model today where businesses are funded upfront by banks in ponzi scheme structures. Enshittification is the natural third stage of these schemes after upfront benefits have been realized.
You burn all bridges going through, make dependent, then become a state run entity (if you are invaluable). Its a common tactic.
The problems people see with google aren't really as they see it at the surface. The issues are the same as the structural problems with central planning in a non-market condition.
There are impossible to solve, intractable problems, and google knows this but the trade-off is in the short-term they get a lot of power and control, and if they are deceptive about it along the way gullible people will be fooled if they don't pay attention until its too late. You see this most clearly demonstrated in their leaked Selfish Ledger.
For those that don't know, its filled with fallacy, and promises that only lead to destructive outcomes and is heavily skewed in marxist thought.
For others its pretty objective, and comes down to whether the use is predicated on the corporate entity being a person, or the entity committing an act.
A fairly common definition of evil is any act that doesn't result in the beneficial and long-term growth of self and others, and as an adjective with regards to people who show a willful blindness towards the consequences of their evil acts (often occurring as a result of many self-violations).
I don't see how what you say would solve or narrow the solution space here given those definitions as they are outcome based.
There are many things with regards to know-how and knowledge that involve knowing both what to do, and what not to do, and you can only really tell if it resulted in destruction or loss after the fact.
Algorithmically to strictly not be evil is to only provide harmless information; but that in itself also leads to maladaptation and development issues which themselves are outcomes where you are evil.
Obeying such a directive would require direct human action on almost everything, and providing just enough information to meet that narrow sliver of beneficial growth but even then you can still mistakenly incur a debt leading you towards the side of the scale of being evil (since we are not omnipotent). All in-actions are themselves actions as well.
While there is precedent for companies like Google, Microsoft, and Cloudflare, to have strong compliance with for example international sanctions, having a Search Engine cripple major publications regularly would be an escalation in my view.
Do you remember how Yahoo went from pretty bad to awfully shit? Guess who was Yahoo's head of search. Spoiler: Prabhakar Raghavan.
I can count the examples of times where non-state ran monopolies benefit the consumer on zero hands.
It's a giant product with thousands of hands in it. Processing it the same way you process other consumer decisions is a mistake.
It is not a monopoly where people don't have a choice. Bing and its derivatives exist, also Yandex and Baidu and several smaller players. Bing is installed by default on every Windows PC. Not only that but the EU is watching Google very closely for antitrust violations, like they did with Internet Explorer back in the days.
Google cannot afford to suck too much or it won't last long. And there is not much lock-in when it comes to search engines.
I used it, it's OK, not really any better overall. Slightly worse in some areas and slightly better in others
The problem seems to be that the incentives set up by Google flooded the web with low quality information
Also, some people probably LIKE some of the low quality information ... but I still think there was a fair amount of mismanagement
“Split-face” would immediately return the horrible accident (don’t Google it, really - and yes, pun intended), “Scientists” would actually return non-editorialized scientists… Microsoft could have been farther and marketed Google as the colourful kindergarten that you give your children access to, whereas journalists and researchers would use Bing. But they folded.
But near the end of the segment "Google with its ... level of intelligence... data... has the ability... to the single best rated product" (a great example of this is trying to find "world's most comfortable chair")
This is unadulterated magic thinking. There is no such thing as "single best" product in any diverse category, especially when you want something so subjective as "comfort". Even if understood literally as "best rated" where theoretically you could have the highest number, there is no source where you could get this number from.
With this fantasy level of Google fu I seriously doubt the author could make an accurate quality assessment of search results besides the "ads take more space now"
And Google Search's past* behaviour was far less toxic.
NB: I've relied all but entirely on DDG search for over a decade. I'm increasingly finding FastGPT from Kagi useful for search-framed-as-question, though the answers have to be carefully vetted. One advantage of Kagi's LLM is that it provides references for its answers making vetting far more straightforward than with, say, OpenAI's LLM.
And my criticism isn't about toxicity as in "ads are bigger", but search result quality, where the level of degradation is an open question. Like with your DDG example: I've tried to switch to it permanently for all search many times, but couldn't because found many instances where Google was just better (think DDG is mostly Bing, which isn't better vs Google)
Now it's mostly really old websites, > ~20 years. Here Google's crawl still seems more robust than DDG's / Bing's. (And yes, DDG leans heavily on Bing.)
Again, I use DDG and retry failed searches via !g bang search. But that's rarely any better.
FastGPT via Kagi is transformational in that I can ask questions and get answers, rather than turning up web pages. Kagi's GPT adds links which reference its sources, which is hugely more useful than ChatGPT in my experience. FastGPT still occasionally hallucinates, but its sources can be much more readily verified. ChatGPT OTOH, hallucinates sources when requested to provide them.
I do have a question regarding the blacklisting, don't you end up having hundreds or even thousands of blacklisted domains? For YouTube I've lost count of the number of channels I asked YouTube to never recommend. It must be way past 200 by now.
I am asking because I get the feeling that there's not much you can do and Google's way is the only thing anyone can do and still be in business. Now how much they rely on ad revenue and how much more they have to compromise is another story. Probably Google should be thinking of including better search to paying Google One customers?
Biggest problem is inertia - that would have to be solved - one way is that Google search gets useless and people have an option to move to something better. I doubt this will happen. Other way is someone builds not just a marginally better product but a 10x better product and people move there - I don't see that happening anytime soon either.
It doesn’t matter what its artificially-defined rules are good. But people would enjoy going there better than on Google, because you’d find the organic pages.
It doesn’t matter that Google would also index them. It’s like the Panamax, it defines rules but others can use those rules to.
The question is how you can be useful enough to the public without being so useful a LLM maker will snap you up
The real problem is the system enables bad actors to do the same, without gaining the Ad label, by gaming the system to outrank legitimate (or free) sources of information.
There's so much irony in that... as with video monetization, Google sucks, except the alternatives don't work or suck worse. Since it's unlikely that Google will have a change of heart about product design, so the only way this can change is if we come up with a better way to search the internet.
Ill say something interesting, since it seems so much needed atm.
When making things I never bother to look at the previous art the way one should. I love figuring things out by myself and usually end up with an inferior version of something fleshed out long ago. This is where the fun begins trying to beat that what exists. It hardly ever happens but I enjoy the process. [ill mark this as section A]
When trying to forge a word and string similarity test there were a lot of approaches to consider. They all worked but nothing is the level of correct that I was looking for. FOO is like FOE but FOO is also like FOOO. If I asked you to calculate how much FOO looks like FOE you could easily slap a number on it. Same for FOOO. It is just impossible to measure likeness as it is subjective but you cant really quantify it either. We could definitely say FOO is 66.666% like FOE and FOOO is 75% like FOO. But then FOOD is also 75% alike while we know FOOO is more like FOO than FOOD.
So I pick some of these "It's nonsense but it works!" and combine them into something that seems to work even better. I had successfully polished the turn it seemed.
I finally read some about popular algos trying to build a site search. One of the hacks was to ignore frequently occurring words. It was pretty funny as the data set I'm trying to search in and the key words I want to search are all frequently occurring. Like 90% of the set is black, 80% is large, 70% is heavy and I specifically need to be able to search "black large and heavy". Something like "to be or not to be" reducing to "".
What google does (and most of the industry) is value short articles with many relevant keywords over long[er] ones. Therefore Section A must be deleted. Much of what I wrote here must be deleted. I have to effectively stop being me and turn into a bland NPC in order to get noticed. Noticed for what exactly?
Maybe I'm a really annoying person unworthy of your attention? If one keeps polishing the turd in that direction eventually your attention will go to objectively annoying people who just want your time and your money.
Then, gradually, day by day, you spend enough time with these fckrs to start thinking it is normal and then, if you are not careful, you become one of them.
Things I've asked it recently include "EVs with 800v battery packs" and "oscilloscopes suitable for hobbyists" and both gave good well-formatted lists in response.
They know that it'll only take a small negative trend in in Ads revenue for investors to freak out.
* 101 Data Protection Tips: How to Protect Your Data (https://www.digitalguardian.com/blog/101-data-protection-tips-how-keep-your-passwords-financial-personal-information-online-safe)
* Protect Your Personal Information and Data (https://consumer.ftc.gov/articles/protect-your-personal-information-and-data)
* Tips for Securing Your Data (https://www.it.northwestern.edu/security/protect-information/secure-data-tips.html)
I have no idea which metrics they're using (and it might be sophisticated enough to justify itself), but I imagine that they could optimize it for revenue per ads or whatever so # of total ads actually are decreased but concentrated on very "high value queries" which users actually care about, so the public perception on # of ads is getting worse and worse. And we know that this kind of "feeling" is very hard to capture.
For reference I did the same search on Google. I do not use Google search by default nor do I ever log in from non-private sessions. I got the same results as you...after one full viewport scroll. Half the SERP is an "AI Overview", below that are some sort of page summaries (not links) for "Secure Your Data", and then finally one full viewport scroll down is the first external link.
DuckDuckGo (my default search engine) isn't doing too much better. The same search gets me three ads before an actual search result. The ads take up 75% of the viewport. Then after only a single outgoing link there's a bunch of links to video results.
To me both SERPs contain mostly garbage. I didn't ask for some AI summary. Nor did I ask for Google to summarize a bunch of pages. I also didn't ask DDG for video results which typically aren't meaningfully different from identified ads.
That was a pain in the ass to search for