It's hard to keep track of the convo on Twitter so I'll comment here.
I think there's something wrong with the data.
The full chart shows a sudden drop of about 9% market share for Google from March going into April 2024. Meanwhile there are significant gains for Bing and Yahoo.
statcounter says they track referrals from search engines (vs. queries directly on the search engine). This suggests to me that there was some change that stopped tracking a significant number of Google referrals, though I don't know enough to know why.
On AI: they do not count moments when a chatbot response references a page, only if you click into the page itself. I think the reactions about ChatGPT/Perplexity eating Google are not entirely correct, considering this + the magnitude of the drop.
> The full chart shows a sudden drop of about 9% market share for Google from March going into April 2024. Meanwhile there are significant [ed:sudden] gains for Bing and Yahoo.
Standard Twitter conversation: blue check guy posts very surprising conclusion. Very surprising conclusion is widely shared. Six inches down the thread, someone points out that this is obviously far too surprising to be real. But if you aren't a member of Twitter all you can see is posts[0].
Google is getting worse and I know I'm not the only one. Reddit is getting full of AI repost bots, meanwhile DDG and Bing seem to be doing ok. What a weird timeline.
My favorite part of DDG is the Bangs that allow me to quickly drop out of DDG when DDG doesn't solve my immediate need. At minimum I can start there, make an attempt then quickly append !g and drop into Google directly. I assume/hope they are using that information to know which searches weren't good enough to improve for the future because it's super useful live feedback.
Check out https://duckduckgo.com/bangs you can dump into pretty much any site, Wikipedia, YouTube, Spotify, Google Translate from the DDG search bar if you have a few favorite bangs memorized. You can put the !<bang> anywhere in the query for it to be picked up as well.
Small correction: you can put it almost anywhere. If you have a quoted search, you need a space between the bang syntax and the quotes, even if the bang syntax comes after.
For example, if you search for ["foobar"!g] (without brackets), you will not be redirected to Google, but you would be if you did ["foobar" !g].
Seems like a data anomaly. With that said, I think it's a good time to go after Google's milkshake:
1. Google clearly lost focus and is rotting. Sundar Pichai makes Steve Ballmer or Brian Krzanich look good by comparison. They're killing search to optimize metrics like "# of searches" that are harmful to long term user value of the platform [1].
Google search has been getting worse for ~5 years, and will not change direction until they get a new CEO. As long as Pichai is at the helm, they'll keep optimizing the KPIs that boost next year's bonus at the expense of making a good service.
2. Kagi does a lot right, but it's run by the kind of lunatic that will spend a significant fraction of the company's budget to buy a t-shirt factory to make t-shirts instead of, you know, making a search engine.
3. Bing exists simply to be a corporate alternative to a competitor. That doesn't make for a good service.
4. DDG <-> bing when it comes to the actual search results. Nothing disruptive in here.
5. LLM based search is nice, but it doesn't fill a lot of the core things people actually search for on a day to day basis. It's a value-add, not a replacement.
Couple of thoughts:
- With generative AI filling the web with noise at an increasing rate, it'll be more important to rely on actual trust of sources rather than the content that's in there. Network/graph based metrics should do better than parsing/embedding the content itself.
- If you're building a search engine service, maybe try to differentiate it for the 21st century? Start by targetting the kind of people that could switch - the same kind of people that already paid for Kagi.
Paying for search is a good idea; it's just that Kagi is run by a loonie. Companies run this way don't tend to last.
Kagi managed to get going with less than 5 engineers. You can bootstrap your project from commoncrawl these days.
> Kagi seems nice, but it's run by the kind of lunatic that will spend a significant fraction of the company's budget to buy a t-shirt factory to make t-shirts instead of
He made a post here explaining that it worked out cheaper that way, and that it hasn’t risked the company. Not that I agree with the decision, but you are painting it in a very bad light.
Kagi is a massive breath of fresh air, putting us, the user, at the helm again. We can choose what to personally up- and downrank, whether we want listicles, long or short site previews, the quick answer LLM is easily available, etc etc etc..
It has been a very very long time since I’ve seen a software product that is such a joyful improvement over all its competitors.
Kagi does almost everything right, except being run by a someone who seems unfit for it.
I'd want someone who is ruthlessly focused on growing the core search product. Focus is about saying no - I don't want a CEO distracted by a bunch of bullshit.
I think it's a good time to eat Kagi's lunch as well.
I'm not a Kagi user, was unaware of the whole tshirt controversy until this thread, and personally I like the kind of lunatic that decides to buy an entire tshirt factory instead of having someone else print them. A man after my own heart
I'm also happily paying for kagi - my point is it's a good time to start a competitor nonetheless.
A startup CEO that gets distracted by T-Shirt factories and all sorts of other shiny squirrels is too unfocused to keep his train on track for the long haul.
> Kagi seems nice, but it's run by the kind of lunatic that will spend a significant fraction of the company's budget to buy a t-shirt factory to make t-shirts
I thought this was a weird joke while reading but apparently it is true. I like it though. Only a lunatic would dare build a paid search engine and actually manage to find users in 2024.
Kagi doesn't have an independent search engine over their own index. Even the Kagi website calls it a search client. Without them creating their own index, I'm not sure it truly qualifies as an independent search engine. Brave Search, which had its full release less than two years ago, does have its own index and is one of the sources that Kagi uses.
They do claim to use an independent Teclis index, but this cannot be where most of their non-news results come from as it is not a general index. From the information available on Teclis, it has a very limited index and currently is strictly purposed for the niche use of finding non-commercial content. Teclis itself doesn't even claim to be fully independent, and sources from the Marginalia non-commercial index. If they were relying on their own index, they wouldn't refer to Kagi as a client and reference the other sources with precedence.
I'd add Brave Search to the list. It's the only Search engine with an independent index that's privacy focused and is actually good!
I've been using it for a few years now and I haven't used Google in a long time. The results are pretty good and it also has pretty nice features compared to Bing/Google/DDG.
The UI is very minimal, not many distractions and there's also LLM search for those who need it.
The Kagi tshirt thing is a generous issue to take up. The owners thoughts on privacy and how it doesn't extent to people deemed criminals is much more problematic.
I think google knows that "search" as in non LLM search is probably dead in the near future, and LLM search other people beat them on. Google search is dead because its already an obsolete product heavily gamed by SEO. They stopped putting effort into it because they're already working on its replacement.
Again, LLM search does some nice things that google search doesn't, but that's also because google search sucks.
Google search has been "llm based" since around 2019 - they've integrated BERT embeddings of your queries for a long time!
Having those embeddings plug into a text generation final layer instead of a retrieval infrastructure isn't somehow magical. ChatGPT only seems better because it's not filled with crap the same way google and bing are.
With that said, search does a lot of things that chatGPT is simply not good at. "Find a hardware store near me". "How does a E15 lightbulb head look like versus the one in my hand", etc.
> 2. Kagi does a lot right, but it's run by the kind of lunatic that will spend a significant fraction of the company's budget to buy a t-shirt factory to make t-shirts instead of, you know, making a search engine.
Now's the time to lobby for Google's unfair cross-screen lock-in to end.
Google shouldn't be the default search engine on Android or Chrome. It's incredibly monopolistic. All users should get the right to choose before defaults are forced upon them.
Their payment of blood money to Apple and Mozilla are also unfair given the fact that they already own all of the other real estate.
It takes less than 30 seconds to change and Chrome is a product _they own_. Take some personal agency. The government doesn't need to hand hold you to change your search engine.
An enormous number of people don't even realize Google isn't the internet.
Defaults are the most powerful force field imaginable at population scale. It's like lead in the water.
Even technical users are dragged down by the second and third order impacts of defaults imposed upon the masses. For instance, Google gets to control how the web works and there's nothing you can do about it.
We should favor robust and healthy competition. Make Google sweat a bit more to earn their place, and give the ecosystem vibrant alternatives that force more rapid innovation.
I had to install Direct X to get a game to run on Linux via WINE. The install modal had an additional message which read something like: Install the Bing bar to enjoy blah blah!! And of course the tick box to install was already ticked.
I felt like I had time warped back to the 90s. So gross. Linux FTW.
I'd like to know the answer to that as well. Kagi still uses Google for some searches, but mixes it in other stuff. However, I do wonder if Googles search is actually fine, if you use it via Kagi or Ecosia, and its just neutered on Googles own site due to prioritizing advertising.
It’s been a while since I compared, but when I did… Bing was almost always better when I did very specific searches, and Google tended to be better at guessing what I wanted to search. Since then, Google has gotten worse at basically everything so… to your question, probably the opposite.
I don't use bing directly but I use DDG which uses bing and I find most of the time its fine but especially for work/science related queries stuff just won't come up and I'll try it on google and get the useful result near the top.
I've used DuckDuckGo from 2012 until 2022, when I switched to Ecosia. Both search engines uses Bing behind the scenes. In my opinion the results have been "good enough" for a long time and as good as Google for at least five years. In the later years Googles results haven't exactly gotten worse, but the layout of their result pages have been degrading fast. So while Google can still find what you're looking for they are hiding those results in favor of showing more ads and trying to keep you leaving their site.
It is a little surprising what Google still maintain such a high market share, but that's momentum and force of habit for you.
It's hard to keep track of the convo on Twitter so I'll comment here.
I think there's something wrong with the data.
The full chart shows a sudden drop of about 9% market share for Google from March going into April 2024. Meanwhile there are significant gains for Bing and Yahoo.
statcounter says they track referrals from search engines (vs. queries directly on the search engine). This suggests to me that there was some change that stopped tracking a significant number of Google referrals, though I don't know enough to know why.
On AI: they do not count moments when a chatbot response references a page, only if you click into the page itself. I think the reactions about ChatGPT/Perplexity eating Google are not entirely correct, considering this + the magnitude of the drop.