One thing I find annoying is that they still return results that are those sites that seem to register a load of terms that all point to the same page. You see this with telephone numbers, song lyrics etc. where the result looks like "Lyrics for Stairway to Heaven" but you click through and there is no content, just a page that says "Upload some lyrics to this song".. etc.
These sites should be heavily penalised for click-baiting and they have been doing it for years.
I've found my result quality has gone up dramatically now that I use extensions allowing me to block domains from google search results. It seems silly that I ought have to do this, but Google has finally become useful again as a result.
Google used to let you do this itself when you were logged in. I never understood why they removed that; surely the information on what domains were deemed unwanted was valuable!?
Are these client-side only extensions that are manipulating the DOM, or are they somehow feeding parameters to google to exclude domains (by e.g., automatically suffixing a series of "-site:blah.com"s to the query)?
I worked on a search engine at a startup that did exactly this, you could up and downvote each result. The main feature was that we essentially "sharded" the search engine so it could be embedded on different sites and give different results based on each community. So a search for "casting" on a fishing website would give different results than the same search on a metallurgy site, as voted by each community. We could also learn passively by watching which links users clicked on and where they "disappeared" from the search engine - presumably the last clicked link was the result they were looking for.
Google did copy the voting feature on their results page briefly but abandoned it. [0] This was back in 2006-7. We learned the hard way that it's pretty much impossible to compete with Google in search even when you're innovating. They either copy you, or can just blackhole you out of existence.
The lack of an ability to vote on search results seems like a baffling omission. Normally Google loves to crowdsource their work, but for this one area where we would actually want it, they decide that they know better when they clearly don't.
"These sites should be penalised for click-baiting and they have been doing it for years."
If SEO works and a result appears closer to result #1 in the SERPs, then the "true", non SEO-assisted result it is displacing would appear further from result #1. Apply this across the board and what we have are many, many non-SEO results that are pushed down in Google's ranking. No one is "penalising" these pages, however they suffer visibility problems because they have not engaegd in SEO. The incentives created by Google's secretive ranking system and online advertising commercial focus are perverse or at least in conflict with the user's goals. Google discourages and even prevents any user from looking at results that were hits but were not ranked high. Pages that may not succombed to the the influence of such incentives may "disappear".
What if a user understands this and wants ignore the Google ranking system. What if the user wants to see the true, non-SEO results. Google actively limits the user's ability to see those displaced results. For example if a user searches for a common term, such as "example", she will not be able to view more than 200-300 results. Elsewhere in this thread someone also noted even with a paid API, Google limits users to 1000 results. If the user wants to the see the full range of pages that have hits for the word "example", she cannot do so. If the user would like to perform a single search for all pages containing the term "example" and then sort by some other objective criteria such as alphabetical by domainname, date, page size, etc., she cannot do so.
Under Google's model of the web, pages that do not acquiesce to an online advertising company's secretive ranking system may become nondiscoverable, despite the fact that they may indeed match the user's query. Computers assist us in searching through data but "relevance" is ultimately decided by the user. That is why we can have HN threads that claim search result quality is declining. Though they may be slower, humans can determine relevance better than any computer. From the disclosures of Matt Cutts and others we know that humans are involved in Google's ranking implementation. Penalties are used. The search process is not 100% math/computer-based. However, in Google's model of the web, filtering results is the exclusive domain of the online advertising company and only the humans on its payroll, not the user performing the search. There is no option to disable the online advertising company's "assistance" in filtering.
Ironically, my work on my own search engine has led me to be a bit more patient with Google's problems. At least I think I understand them better. Search engines fail in weird ways.
I think in part that Google just has gotten a spectacularly confusing failure mode. If it can't find good matching contents, it starts second-guessing your query and producing other results, which makes you think it's not even considering what you entered. It may even be "better" in the sense that it's more likely to return at least something relevant, but in practice it's bad UX because it's so unintuitive what's happening. It's probably one of those unfortunate optimizations that are invisible when they work and frustrating when they don't.
There is so much stuff on the Internet it's easy to start thinking there is guaranteed to be good results for any search, and that just doesn't seem to be the case. Especially with highly specified searches with 6-8 terms, you quickly enter the domain where you're reasonably unlikely to find an exact match.
> I think in part that Google just has gotten a spectacularly confusing failure mode. If it can't find good matching contents, it starts second-guessing your query and producing other results, which makes you think it's not even considering what you entered.
This is probably part of it but not the whole explanation:
Try to search on Google for:
slack ngrok
When I and others did earlier today there were a number of pages that contained the words including from the slack.com domain.
The top result however was a page that didn't contain ngrok at all.
I saw a specialist at another search engine comment that it was because it was a very popular result (at least thats what I read into it).
Here's my problem with Google: they are either just really bad at QA or they don't care or they consistently overestimate their dumb AI and underestimate me.
I'm fed up.
Not including pages that doesn't contain the search terms or anything similar isn't hard when there are multiple good results at the exact same domain / pagerank, is it?
Given the term "ngrok", I can't say I blame them/it. It looks like a typo to me. I imagine for well over 90% of people searching, it would be a typo. Perhaps, it is a common typo?
I've taken this possibility into account after the numerous recent threads about this, but I don't think it holds up. I'm very often able to find exactly what I'm looking for after 15 minutes of massaging search terms on 3-4 different search engines. The reason it feels like search has gone to hell, for me, is that for 20 years I took for granted that I could type the first terms that came into my mind into one search engine (Google) and the result I wanted was the first result.
Outside of programming-related topics, anything I search returns pages of pop psychology listicles or news articles. Since I am literally never looking for pop psych listicles, Google (and, to be fair, the other search engines as well) has become a lot less usable.
I agree that the open web has deteriorated, with crap drowning out real content. But I maintain that Google et al have failed, or been beat. The content is there, they just can't find it and/or rank it anymore.
Yes, the more popular terms you search for, the more muddy the waters get.
If you search for anything related to sexuality and psychology, the results are littered with sites that were squatted for serving ads (i.e. no relevant content, just ads), poor quality articles with very low quality content (e.g. poorly formed Quora questions with no expert answers).
As with anything, you have to know what search terms for a given subject are good. For example, you'll get more objective answers the more academic sounding you are, because the fewer people have tried to occupy those search terms.
I agree with your experience - to me it also seems I used to enter a search phrase and get back a page of results that exactly match it, these days, I often get things that only vaguely thematically relate to my search phrase instead. Sometimes I would change a word in my search phrase or add an extra word to make the intent much more specific and still get essentially the same search results!!! Terrible, just terrible.
If I'm searching for a development issue, and my first results page contains 2-3 sites that 1-1 copy stackoverflow/github issues, then Google has failed. I doubt those can be more authorative than the original sources
Even programming related topics now return heavily SEOd sites with instead of high quality programming content. Often the SEOd content is not completely awful for these searches, but it’s usually not great & never as good as the best sites.
> I'm very often able to find exactly what I'm looking for after 15 minutes of massaging search terms on 3-4 different search engines.
If you are adjusting your query, then you are going to get different results, possibly including ones that contain the information you want (but not your original search terms).
We've been tweaking search terms to find what we were looking for since day one, what's changed the most it's that it's gotten far less likely that you'll realize you need to do this.
I think this is one of the real drawbacks of ML-algorithms, their failure modes are completely incomprehensible. Dumb algorithms we can grok, and learn to help along the way when it doesn't work. There is really no point where it will always work.
Either google failed, or the pop science listicles have won ... they're not passively being indexed, sorry you found them by mistake. They actively and aggressively create completely artificial content, tailored not at humans but at google, so that they can push the latest stupid ad for some telephone company on you.
It's like bitcoin if you want - they compete for something so useless the entire concept becomes a huge waste of time. Search engine SEO is like hashrate-dependent token mining: the only people who win are the farmers at the cost of burning their entire ecosystem.
> There is so much stuff on the Internet it's easy to start thinking there is guaranteed to be good results for any search, and that just doesn't seem to be the case.
I'm increasingly of the opinion that Google (the advertising engine) has destroyed Google (the search engine), by the two step process of making it profitable to produce blogspam then forcing search to remove blogspam - and a lot of the useful content has gone out with that bathwater.
Not to mention the rise of unsearchable platforms. Google can't search inside Discord.
> I'm increasingly of the opinion that Google (the advertising engine) has destroyed Google (the search engine), by the two step process of making it profitable to produce blogspam then forcing search to remove blogspam - and a lot of the useful content has gone out with that bathwater.
If you're looking for someone to blame google seems like the wrong party here. Surely the people abusing the system (i.e. blog spam) should at least share a good chunk of the blame.
Can anyone, even in theory? Are there open APIs to all systems of discord? Does Discord have one? Wouldn't that open up all of these systems to systematic classification of all users? Also, is there a web link you can construct that will open up e.g. Discord's desktop app when you click it?
On Amazon I searched in "CDs & Vinyl" for the band "The Birdstones". It showed me results in books.
Criminy. Amazon does this all the time. If I wanted a book, I'd have searched under "Books". I am not confused about the difference between "CDs" and "Books".
What's the good of having categories if they're completely ignored?
I now avoid Amazon for this reason. I'll search for something like "4k HDMI capture card" and half the results are 1080p (and don't say 4k or a synonym anywhere) and a tenth do DisplayPort. I end up having to open every linked product and use my borwser search to confirm that my requirements are actually present.
Same with any sort of requirement like "plastic" a colour...
It just wastes my time so I shop elsewhere.
I think a much better UX would be saying: No results but consider searching for "4k capture card" or "HDMI capture card".
Some funny example is "Eclipse": Do you mean Eclipse the IDE? Or the vampire movies? Maybe you mean the astronomical event? or the Nissan Eclipse car? Or you wrote wrong ellipse?
I'm fine with Google not reading my mind at the first try, but at least offer me alternatives and exact text matching. For example since a couple of years looking for phones or symbols is completelly broken.
And I really admire how Google is able to guess the encoding of the websites, detect what is text, do language detection and dropping all the porn. Writting a crawler that actually works is HARD.
When I search for "databricks series b valuation" in Google (from Argentina, using Google.com in English) result #6 is:
"Python get value from database - Büro Jorge Schmidt", which judging by title and preview seems to be a Python + MySQL tutorial. It returns a 403 error and might be a hacked site, since the home page is for a graphic design studio in Munich.
Result #8 is something similar:
"Intellij flatten packages - Músicos de Viaje". This is definitely a hacked site (from Spain, apparently) that redirects me somewhere else.
Result #10:
"How to calculate tax percentage in sql query". Another hacked site, this time for an evangelical church from Brazil.
Now... how can Google think that any of these sites are relevant? Even if it doesn't realize the pages are hacked... even its crawler has been fed content that included the keywords... :
A - The sites themselves don't match the query at all.
B - No legit site about the subject would link to these sites.
C - The results themselves (title, url, preview), as Google shows them, have nothing to do with the search!
I just tried that search, all of the results look relevant and I definitely don't get any of the results you are getting.
I wonder if you have some malware that is hijacking the results? I once had some malware (chrome extension) that was corrupting my search results. It was surprisingly difficult to remove (given that it was a chrome extension...).
No, I get the same results on Safari on iOS (iPhone) so I have to eliminate the possibility of malware.
Google results are personalized, based on location, search history, etc. The fact that I'm in Argentina has been adding a lot of noise to results on searches where my location is not relevant at all.
In this case, I suspect that Google thinks these hacked sites with developer target content are relevant to me, because of my regular search history.
Just tried that search in English. First result is "The data- and AI-focused company has secured a $1.6 billion round at a $38 billion". And if you click the search box you get the first search option as databricks valuation history where the first result is funding every round.
Google has been so much better for search for me than other search engines. Atleast for what I search for programming, news etc.
- I checked the cached (by Google) copies of the hacked pages and they include mentions of a "Databricks SQL Connector". So if I search for "Databricks" Google thinks "it must be a programming thing".
- If I now search for "databricks series a valuation" I don't get the spam results, for some reason. I think that if I repeat the search Google produces the exact same results... but internally, since I first searched, it might have realized that those sites were not good.
That being said, I have issued queries in the past that I have just found absolute walls of malicious results. I haven't really invested much attention span in entertaining this sort of thing so I just move on and modify my query, but I'll keep an eye out for it going forward.
The article mostly talks about IA (instant answers) which are notoriously hard. The recent advances in machine learning have made the technology more approachable, so startups like Kagi Search (disclaimer: founder) can also leverage latest advances in NLP and compete on this ground.
Both engines use the same article for source, but Google completely misses the context.
These examples show that a search startup has a chance to go neck-to-neck with Google and compete even in technology as sophisticated as instant answers. We invested considerable resources in the Kagi Search AI capabilities, discussed in some detail here https://kagi.ai/last-mile-for-web-search.html
What is mind boggling though from a product management perspective is that Google had nearly a decade head start and a cash purse of hundreds of billions of dollars to get this right.
To be fair, it is likely that the vast majority of queries are answered correctly, but only the outliers get the public attention. Also Kagi is not without its own share of silly mistakes too, but just being able to be considered in the same basket as Google is already a huge thing for us.
My favorite is, given that I have a baby and I am a trained scientist and I live in the US, I find myself converting milliliters to fluid ounces a lot.
Right now the Google Assistant will correctly transcribe the request to a Google search... Only for the search to interpret “ml” as “miles” and, faced with the discrepancy between the length and a volume, cube the miles. So I am expecting an amount that is like 1 oz because I am converting like 30 mL... and I instead get 4 quadrillion ounces (exact number is 1 mi³ = 140,942,994,870,857 + 1/7 oz because of course it's got that extra 7th in there what were you expecting from our ridiculous US system, haha).
> No one at Google is responsible for these half baked and largely irrelevant widgets or wants to stake their career fixing them.
You're just ... wrong about this. There's an entire team of dozens of people (maybe hundreds now) focusing on this specific web answer feature. I personally worked on the team (not this feature, though).
I don't understand why people say things they know nothing about.
> The article mostly talks about IA (instant answers)
Not really. TFA is discussing the search results. If google/bing want to put instant answers then that is what will get judged. If your ML/AI is not good enough yet to provide natural language answers, don't make it the most prominent part of the search results.
I think that Google is optimizing for the "average user" to the detriment of power users such as the HN crowd. Most people treat Google as an internet oracle and send queries like "how do I do X" while power users will search for keywords. One example of this optimization is the automatic answer boxes that show up for certain questions, which are wrong disturbingly often or don't include important details.
The average use absolutely does this. I see this with family members and friends. Most just type in full questions.
From my experience the best way to get good results is to start typing keywords for yourr question and then creating the query based on the autocomplete results. If I notice I don't get autocompletion for a certain query I'll restructure it until I do. This has proven very useful in providing good results.
For technical stuff, using the quotation marks is almost essential.
I miss Altavista. I could generally either find exactly what I was looking for inside of three iterations of refining the search, or find that it wasn't to be found.
I still just want a blazing fast full text search of the reachable WWW that understands regexes and a basic predicate calculus. Unfortunately the overhead and small potential user base means that under the current regime such a thing will never be made.
Speaking of, if any government actually wants competition, they don't need to break up Google, they just need to force them to offer full access to their cache and compute at some reasonable rate, much like how the ILECs were made to carry the CLECs' traffic.
This. And let's not forget that is the "average user" that mostly naively clicks on ads, not the power user. And selling ads is still Google core business.
Looks like Google is slowly turning into a big nigerian scam.
I think this is overall a good thing. Power users have trained their behaviour to work in the way that simple systems can deal with. While average users ask the question exactly how they would ask another human. Google has now reached a level where it works best when you deal with it in a natural and human level.
There is nothing actually better about the way we originally used search engines, it was just required at the time.
No, I disagree. With keyword search, I am confident that eventually, with enough included and excluded terms, I will find what I’m looking for.
With natural language search, sometimes it works great, but it’s a crapshoot and when you don’t get the results you need you’re stuck.
Several times in recent memory Google has returned results so bad, I completely gave up searching. Most recently was when i was trying to look up a Windows 11 BSOD error code (where even pasting the error code verbatim only brought up pages of garbage sites with no useful technical information).
For some queries, being able to ask off the top of your head without thinking is good. Think 'what day of the week was June 14th, 2002?' or 'Who is the mayor of Los Angeles?' For quick questions with clear answers, the current system is a huge advantage over what we had before.
For other, more complicated queries, the act of composing your search and considering your keywords, etc. is a step in the process that helps a searcher mentally understand the results they're going to receive along with what they mean. Having to stop and consider makes you aware that you're working within a system and its constraints, which makes it more suitable for questions that are complex or not socially settled.
Instant answers (IA) caused a shift in the way contents are written. Content optimized for IA tend to be repetitive and shallow. Viewing content written for IA is a frustrating experience and these tend to dominant the result page now.
The reason you are seeing a lot of that sort of content is because Google is looking for that sort of content, in point because of Google's peculiarities, but also because of how refined the art of black hat SEO has become.
Meaningful websites still exist. The bulk of the content on the Internet is older than IA, and it's still out there (not that you can find it with Google).
Hey, free advertising for you and your search engine:
Marginalia search, by punishing ad- and tracking-heavy pages and by being strict in how it interprets my queries sometimes surfaces better results than Google and DDG.
In particular I have found great resources about Linux partitioning and git usage after giving up mainstream search engines and trying them in marginalia.
That says quite something about how badly broken the situation is given that marginalia is one person and a tower pc in a living room.
I keep getting reminded about a Linux quote on how they managed to go forward by studying the latest 20 years of OS research and throwing it all away :-)
No idea if you're right or not, but I think it's an interesting take. It might be because of SEO, it might be more information in walled gardens, it might be the death of the personal web page.
Both Google search and the internet are changing, as are our perceptions, so it's really hard to say why search quality is better or worse.
Whether search quality is deteriorating really depends what we expect from a search engine. Many of us on HN use search engine as a way to look for web pages with specific words. Search engines did start out performing word searches, but somewhere along the way, they started trying to answer questions. Many of the complaints such as Google dropping search terms are actually the result of the search engine trying to answer questions instead of performing keyword searches. Search engines these days are doing well for the easy questions, but of course they completely misunderstand questions with more nuances.
A bit off topic: Some of the search queries in the article used Google like a keyword matching tool, while other queries in the article were using Google to answer questions. That's because we only have one search box trying to do everything. Would we be better served if we have a checkbox to tell the search engine that we just want to perform word searches?
It's definitely deteriorating, and the worst part is that it completely ignores quotes if it thinks you meant something else, and shows the results for what it thinks you want. Completely useless in a lot of cases
I have an Intel Realsense camera, which sometimes reports the error "Failed to recconect" (there being a typo in the drivers) [1] - that's a pretty unique error, so in combination with the product name that should be a very easy keyword search, right?
But no, when I search for realsense "failed to recconect" Google returns pages that contain neither realsense nor recconect [2]. They offer me a supreme court opinion, a review of a car dealership, and a facebook church service.
Correcting the spelling of a query is one thing - but also completely ignoring other keywords? I can see why there are so many people posting about the poor quality of Google's search results.
This has only happened to me twice so far, but it has given me the "Did you mean:" even when using quotes for searches that definitely have results (as I found them with duckduckgo).
I can't remember what they were now
If you search for anything political on Google, you'll notice that the results are clearly slanted in one direction, towards the opinion of a handful of pre-approved news outlets. This leads me to seek alternatives whenever I need neutral sources, for instance Yahoo search.
I wonder if they're getting cash kickbacks from established corporate media outlets for pushing their material to the top of the search results. That would actually be less creepy than if its being done as some kind of information manipulation program.
It's high time Google and other search engines were forced to expose the inner workings of their ranking algorithms to the public, particularly now that they have near-monopoly power in the sector. People should also be able to adjust the dials on the algorithm themselves.
In Australia, Google is being blackmailed into boosting Fairfax, Newscorp and Seven West Media content higher up in their index. It's reached a point where most queries are useless if they contain a word or even a synonym for that word that has been recently used in a major media site owned by those three.
I use Google through a VPN to avoid it. That breaks maps integration.
I think it is related to the fight against "fake news", "hate speech", etc... People don't tolerate a truly neutral search engine, because it will reflect human nature and human nature is not always pretty. I remember the time when Google returned antisemitic websites when searching for "jew", they refused to do anything about it because "jew" is used mostly by antisemites and therefore, an antisemitic website is what people searching for that term most likely want, the search engine did its job. I don't think it will fly today.
So search engines now have to get the "truth", preferably the politically correct one, and since you can't rely on the crowd for that, you have to introduce bias, and "pre-approved news outlets" are the most obvious choice.
Try "what countries are using ivermectin" in google.com and then try Yandex.com. For me, the third site on Google (the kitchen sisters) appears broken and the rest are all some variation of "why ivermectin is bad" articles. Yandex actually answers the question.
I find these responses fascinating as the "clearly slanted" results tend to change direction depending on the political affiliation of the person making the claim! Having said that, I'd love to be proved wrong if you have any evidence to show a particular bias one way or the other?
Search "mass formation psychosis" on Google and DuckDuckGo. This is a trending phrase popularized by a doctor that has been canceled and banned from the mainstream due to his criticism of the world's COVID response.
DDG shows the author's substack as #1 result and is neutral otherwise. The other doesn't even have it on the first SERP, and is overwhelmingly critical.
If you argue that covid response is not politics, I will disagrer strongly.
Search politically controversial topics incognito with a VPN on using google vs yandex vs duckduckgo? Ivermectin, January 6th arrests, BLM protest deaths, VAED, mRNA studies before 2020, Robert Malone, Geert Vanden Bossche. I mean there's an endless list of things you can experiment with.
My experience is that it's slanted whichever side of the easel you're one. The evidence is really clear when you measure results between various search providers, and especially when searching up contentious or controversial topics. So it doesn't really change direction so much as it confirms one particular set of beliefs depending on which search engine you're on. I think this is clearly in Google's disfavour, because people have started to notice and they're actively searching for alternatives to Google in order to avoid it.
As a corollary, search on Google News (as in, browsing to news.google.com and searching there, or !gn via DuckDuckGo) is really bad. The index seems to update really slowly, so breaking events are usually missing entirely, and the grouping of articles into single events is also quite broken.
Non-factual answers are especially interesting, because otherwise reasonable and intelligent people believe in them. Everyone would be better off studying those things and see how they've drawn those conclusions we don't subscribe to, so that we can figure out what we are wrong about.
It's absolutely catastrophic if you are not allowed to draw your own conclusions about things. This is your inalienable prerogative as an adult in a free country. Even at the risk of some people being wrong sometimes, you simply cannot have authorities distributing doctrine and call yourself a democracy.
Readit News
These sites should be heavily penalised for click-baiting and they have been doing it for years.
Google did copy the voting feature on their results page briefly but abandoned it. [0] This was back in 2006-7. We learned the hard way that it's pretty much impossible to compete with Google in search even when you're innovating. They either copy you, or can just blackhole you out of existence.
[0] https://techcrunch.com/2007/11/28/straight-out-of-left-field...
If SEO works and a result appears closer to result #1 in the SERPs, then the "true", non SEO-assisted result it is displacing would appear further from result #1. Apply this across the board and what we have are many, many non-SEO results that are pushed down in Google's ranking. No one is "penalising" these pages, however they suffer visibility problems because they have not engaegd in SEO. The incentives created by Google's secretive ranking system and online advertising commercial focus are perverse or at least in conflict with the user's goals. Google discourages and even prevents any user from looking at results that were hits but were not ranked high. Pages that may not succombed to the the influence of such incentives may "disappear".
What if a user understands this and wants ignore the Google ranking system. What if the user wants to see the true, non-SEO results. Google actively limits the user's ability to see those displaced results. For example if a user searches for a common term, such as "example", she will not be able to view more than 200-300 results. Elsewhere in this thread someone also noted even with a paid API, Google limits users to 1000 results. If the user wants to the see the full range of pages that have hits for the word "example", she cannot do so. If the user would like to perform a single search for all pages containing the term "example" and then sort by some other objective criteria such as alphabetical by domainname, date, page size, etc., she cannot do so.
Under Google's model of the web, pages that do not acquiesce to an online advertising company's secretive ranking system may become nondiscoverable, despite the fact that they may indeed match the user's query. Computers assist us in searching through data but "relevance" is ultimately decided by the user. That is why we can have HN threads that claim search result quality is declining. Though they may be slower, humans can determine relevance better than any computer. From the disclosures of Matt Cutts and others we know that humans are involved in Google's ranking implementation. Penalties are used. The search process is not 100% math/computer-based. However, in Google's model of the web, filtering results is the exclusive domain of the online advertising company and only the humans on its payroll, not the user performing the search. There is no option to disable the online advertising company's "assistance" in filtering.
I think in part that Google just has gotten a spectacularly confusing failure mode. If it can't find good matching contents, it starts second-guessing your query and producing other results, which makes you think it's not even considering what you entered. It may even be "better" in the sense that it's more likely to return at least something relevant, but in practice it's bad UX because it's so unintuitive what's happening. It's probably one of those unfortunate optimizations that are invisible when they work and frustrating when they don't.
There is so much stuff on the Internet it's easy to start thinking there is guaranteed to be good results for any search, and that just doesn't seem to be the case. Especially with highly specified searches with 6-8 terms, you quickly enter the domain where you're reasonably unlikely to find an exact match.
This is probably part of it but not the whole explanation:
Try to search on Google for:
When I and others did earlier today there were a number of pages that contained the words including from the slack.com domain.The top result however was a page that didn't contain ngrok at all.
I saw a specialist at another search engine comment that it was because it was a very popular result (at least thats what I read into it).
Here's my problem with Google: they are either just really bad at QA or they don't care or they consistently overestimate their dumb AI and underestimate me.
I'm fed up.
Not including pages that doesn't contain the search terms or anything similar isn't hard when there are multiple good results at the exact same domain / pagerank, is it?
It is the top result on Bing as well. It probably shows up in this spot on both search engines because it's prominently linked on https://ngrok.com/.
Given the term "ngrok", I can't say I blame them/it. It looks like a typo to me. I imagine for well over 90% of people searching, it would be a typo. Perhaps, it is a common typo?
Outside of programming-related topics, anything I search returns pages of pop psychology listicles or news articles. Since I am literally never looking for pop psych listicles, Google (and, to be fair, the other search engines as well) has become a lot less usable.
I agree that the open web has deteriorated, with crap drowning out real content. But I maintain that Google et al have failed, or been beat. The content is there, they just can't find it and/or rank it anymore.
If you search for anything related to sexuality and psychology, the results are littered with sites that were squatted for serving ads (i.e. no relevant content, just ads), poor quality articles with very low quality content (e.g. poorly formed Quora questions with no expert answers).
As with anything, you have to know what search terms for a given subject are good. For example, you'll get more objective answers the more academic sounding you are, because the fewer people have tried to occupy those search terms.
If you are adjusting your query, then you are going to get different results, possibly including ones that contain the information you want (but not your original search terms).
We've been tweaking search terms to find what we were looking for since day one, what's changed the most it's that it's gotten far less likely that you'll realize you need to do this.
I think this is one of the real drawbacks of ML-algorithms, their failure modes are completely incomprehensible. Dumb algorithms we can grok, and learn to help along the way when it doesn't work. There is really no point where it will always work.
It's like bitcoin if you want - they compete for something so useless the entire concept becomes a huge waste of time. Search engine SEO is like hashrate-dependent token mining: the only people who win are the farmers at the cost of burning their entire ecosystem.
I'm increasingly of the opinion that Google (the advertising engine) has destroyed Google (the search engine), by the two step process of making it profitable to produce blogspam then forcing search to remove blogspam - and a lot of the useful content has gone out with that bathwater.
Not to mention the rise of unsearchable platforms. Google can't search inside Discord.
If you're looking for someone to blame google seems like the wrong party here. Surely the people abusing the system (i.e. blog spam) should at least share a good chunk of the blame.
Can anyone, even in theory? Are there open APIs to all systems of discord? Does Discord have one? Wouldn't that open up all of these systems to systematic classification of all users? Also, is there a web link you can construct that will open up e.g. Discord's desktop app when you click it?
Criminy. Amazon does this all the time. If I wanted a book, I'd have searched under "Books". I am not confused about the difference between "CDs" and "Books".
What's the good of having categories if they're completely ignored?
Same with any sort of requirement like "plastic" a colour...
It just wastes my time so I shop elsewhere.
I think a much better UX would be saying: No results but consider searching for "4k capture card" or "HDMI capture card".
Showing results from All Departments
No results for The Birdstones in Music, CDs & Vinyl
I think this widening of search scope is not unreasonable.
Deleted Comment
I'm fine with Google not reading my mind at the first try, but at least offer me alternatives and exact text matching. For example since a couple of years looking for phones or symbols is completelly broken.
And I really admire how Google is able to guess the encoding of the websites, detect what is text, do language detection and dropping all the porn. Writting a crawler that actually works is HARD.
"Python get value from database - Büro Jorge Schmidt", which judging by title and preview seems to be a Python + MySQL tutorial. It returns a 403 error and might be a hacked site, since the home page is for a graphic design studio in Munich.
Result #8 is something similar:
"Intellij flatten packages - Músicos de Viaje". This is definitely a hacked site (from Spain, apparently) that redirects me somewhere else.
Result #10:
"How to calculate tax percentage in sql query". Another hacked site, this time for an evangelical church from Brazil.
Now... how can Google think that any of these sites are relevant? Even if it doesn't realize the pages are hacked... even its crawler has been fed content that included the keywords... :
A - The sites themselves don't match the query at all.
B - No legit site about the subject would link to these sites.
C - The results themselves (title, url, preview), as Google shows them, have nothing to do with the search!
I wonder if you have some malware that is hijacking the results? I once had some malware (chrome extension) that was corrupting my search results. It was surprisingly difficult to remove (given that it was a chrome extension...).
Google results are personalized, based on location, search history, etc. The fact that I'm in Argentina has been adding a lot of noise to results on searches where my location is not relevant at all.
In this case, I suspect that Google thinks these hacked sites with developer target content are relevant to me, because of my regular search history.
Do you geolocate in Argentina?
Google has been so much better for search for me than other search engines. Atleast for what I search for programming, news etc.
Also, results are personalized. Not everyone sees the same.
If I do an incognito search (not logged in) I get results without trash.
- I checked the cached (by Google) copies of the hacked pages and they include mentions of a "Databricks SQL Connector". So if I search for "Databricks" Google thinks "it must be a programming thing".
- If I now search for "databricks series a valuation" I don't get the spam results, for some reason. I think that if I repeat the search Google produces the exact same results... but internally, since I first searched, it might have realized that those sites were not good.
That being said, I have issued queries in the past that I have just found absolute walls of malicious results. I haven't really invested much attention span in entertaining this sort of thing so I just move on and modify my query, but I'll keep an eye out for it going forward.
Deleted Comment
To give just a few examples:
Query 1: how many stars in the usa flag
Google: https://cln.sh/63sVzh
Kagi: https://cln.sh/bFEHsD
Pretty surprising that Google would get something like this wrong.
Query 2: when did moon explode
Google https://cln.sh/fUhdJS
Kagi https://cln.sh/5wDvXG
Both engines feature the same article but for some reason Google decides this is not fiction, and gives a (wrong) answer.
Query 3: do most rabbits have short or long ears
Google: https://cln.sh/JuOeqq
Kagi: https://cln.sh/BkZi6O
Both engines use the same article for source, but Google completely misses the context.
These examples show that a search startup has a chance to go neck-to-neck with Google and compete even in technology as sophisticated as instant answers. We invested considerable resources in the Kagi Search AI capabilities, discussed in some detail here https://kagi.ai/last-mile-for-web-search.html
What is mind boggling though from a product management perspective is that Google had nearly a decade head start and a cash purse of hundreds of billions of dollars to get this right.
To be fair, it is likely that the vast majority of queries are answered correctly, but only the outliers get the public attention. Also Kagi is not without its own share of silly mistakes too, but just being able to be considered in the same basket as Google is already a huge thing for us.
Right now the Google Assistant will correctly transcribe the request to a Google search... Only for the search to interpret “ml” as “miles” and, faced with the discrepancy between the length and a volume, cube the miles. So I am expecting an amount that is like 1 oz because I am converting like 30 mL... and I instead get 4 quadrillion ounces (exact number is 1 mi³ = 140,942,994,870,857 + 1/7 oz because of course it's got that extra 7th in there what were you expecting from our ridiculous US system, haha).
Perhaps everyone who understood the codebase has left?
No one at Google is responsible for these half baked and largely irrelevant widgets or wants to stake their career fixing them.
You're just ... wrong about this. There's an entire team of dozens of people (maybe hundreds now) focusing on this specific web answer feature. I personally worked on the team (not this feature, though).
I don't understand why people say things they know nothing about.
The third query is wrong for me too, though.
Not really. TFA is discussing the search results. If google/bing want to put instant answers then that is what will get judged. If your ML/AI is not good enough yet to provide natural language answers, don't make it the most prominent part of the search results.
Same results as you on the third one though.
From my experience the best way to get good results is to start typing keywords for yourr question and then creating the query based on the autocomplete results. If I notice I don't get autocompletion for a certain query I'll restructure it until I do. This has proven very useful in providing good results.
For technical stuff, using the quotation marks is almost essential.
I still just want a blazing fast full text search of the reachable WWW that understands regexes and a basic predicate calculus. Unfortunately the overhead and small potential user base means that under the current regime such a thing will never be made.
Speaking of, if any government actually wants competition, they don't need to break up Google, they just need to force them to offer full access to their cache and compute at some reasonable rate, much like how the ILECs were made to carry the CLECs' traffic.
Looks like Google is slowly turning into a big nigerian scam.
There is nothing actually better about the way we originally used search engines, it was just required at the time.
With natural language search, sometimes it works great, but it’s a crapshoot and when you don’t get the results you need you’re stuck.
Several times in recent memory Google has returned results so bad, I completely gave up searching. Most recently was when i was trying to look up a Windows 11 BSOD error code (where even pasting the error code verbatim only brought up pages of garbage sites with no useful technical information).
P: "Pfft. So, what? You think the current state of search is good?! Having to type in keywords instead of just asking a question in normal language?"
Me: "... is this a trick question?"
For some queries, being able to ask off the top of your head without thinking is good. Think 'what day of the week was June 14th, 2002?' or 'Who is the mayor of Los Angeles?' For quick questions with clear answers, the current system is a huge advantage over what we had before.
For other, more complicated queries, the act of composing your search and considering your keywords, etc. is a step in the process that helps a searcher mentally understand the results they're going to receive along with what they mean. Having to stop and consider makes you aware that you're working within a system and its constraints, which makes it more suitable for questions that are complex or not socially settled.
Not all information is the same.
Instant answers (IA) caused a shift in the way contents are written. Content optimized for IA tend to be repetitive and shallow. Viewing content written for IA is a frustrating experience and these tend to dominant the result page now.
Meaningful websites still exist. The bulk of the content on the Internet is older than IA, and it's still out there (not that you can find it with Google).
Marginalia search, by punishing ad- and tracking-heavy pages and by being strict in how it interprets my queries sometimes surfaces better results than Google and DDG.
In particular I have found great resources about Linux partitioning and git usage after giving up mainstream search engines and trying them in marginalia.
That says quite something about how badly broken the situation is given that marginalia is one person and a tower pc in a living room.
I keep getting reminded about a Linux quote on how they managed to go forward by studying the latest 20 years of OS research and throwing it all away :-)
Both Google search and the internet are changing, as are our perceptions, so it's really hard to say why search quality is better or worse.
A bit off topic: Some of the search queries in the article used Google like a keyword matching tool, while other queries in the article were using Google to answer questions. That's because we only have one search box trying to do everything. Would we be better served if we have a checkbox to tell the search engine that we just want to perform word searches?
Getting this wrong is probably why people think we need more than definitive assertions from people operating from subjective impression.
Changing a letter in the query, it says, right at the top:
But no, when I search for realsense "failed to recconect" Google returns pages that contain neither realsense nor recconect [2]. They offer me a supreme court opinion, a review of a car dealership, and a facebook church service.
Correcting the spelling of a query is one thing - but also completely ignoring other keywords? I can see why there are so many people posting about the poor quality of Google's search results.
[1] https://github.com/IntelRealSense/librealsense/blob/5ff27fca... [2] https://imgur.com/a/okYV5V2
I rarely see "No results found".
That is a completely valid result in my opinion.
I often see pages that doesn't contain my search string.
Sometimes is is something trivial like
query: "something about x y and z"
result contains: "... something about x. Y and Z however..."
which is understandable.
Often though I have no clue how a result got there. Maybe it is linked to using a certain text (link bombing)? I don't know.
Deleted Comment
Deleted Comment
It's high time Google and other search engines were forced to expose the inner workings of their ranking algorithms to the public, particularly now that they have near-monopoly power in the sector. People should also be able to adjust the dials on the algorithm themselves.
I use Google through a VPN to avoid it. That breaks maps integration.
So search engines now have to get the "truth", preferably the politically correct one, and since you can't rely on the crowd for that, you have to introduce bias, and "pre-approved news outlets" are the most obvious choice.
DDG shows the author's substack as #1 result and is neutral otherwise. The other doesn't even have it on the first SERP, and is overwhelmingly critical.
If you argue that covid response is not politics, I will disagrer strongly.
It's absolutely catastrophic if you are not allowed to draw your own conclusions about things. This is your inalienable prerogative as an adult in a free country. Even at the risk of some people being wrong sometimes, you simply cannot have authorities distributing doctrine and call yourself a democracy.
Deleted Comment