How Kagi finally let me lay Google Search to rest

I’m a Kagi convert, having switched from Google Search in July of 2022 and never looking back. I’m also a vocal supporter of Kagi, frequently mentioning it in my newsletter and encouraging people to give Kagi a try whenever the opportunity arises.

For those unfamiliar, Kagi is a search engine with a novel¹ business model: end users pay for a service. Instead of monetizing users through advertising and affiliate marketing (like every other search engine), Kagi charges a monthly fee.

I’m both a user and vocal supporter of Kagi, and I also have big thoughts and opinions. Both on the state of search today, where it’s going in the future, and why it’s important to support companies like Kagi. Additionally, I was blown away by the quality of Kagi’s search results, which is what first encouraged me to make the switch, and I wanted to dive into the details in case anyone reading is on the cusp of trying this new search engine.

This post is not an ad; I’m just a happy customer. And I’ll gladly pay for products I like, and actively hope others support these products, too, so that the companies that make them can survive.

With that said, let’s kick things off with a little background on Internet Search as it exists today.

Wow, Google Search sucks

It may be easy to miss, but Google Search is dying. Search results are dominated by ads SEO-optimized junk. More and more, people append words like “reddit” to the end of their search to try and bubble up any semblance of a useful response.

According to the author above blog post (with whom I agree): “serving ads creates misaligned incentives for search engines.” It’s impossible to both provide the best search results and try to optimize for the highest amount of ad-clicks. And if you’re a public company with an ad-based business model, you are legally required to optimize for the latter. Ads inherently create misaligned incentives.

But Google continues to dominate as the most-used search engine worldwide. I think part of this has to do with the fact that it’s so ingrained into people’s lives that its flaws go unnoticed or overlooked. People just amend their behavior (like adding “reddit” to the end of a search, or immediately scrolling through pages of SEO-junk to get to a recipe) to compensate for Google’s shortcomings.

Google reigns supreme, too, in no small part because historically all its competitors all suck even more. Let’s take a quick look at the mainstream alternatives. As of August 2023, US desktop search engine market share is as follows:

1. Google (78.96%)
2. Bing (14.42%)
3. Yahoo (3.89%)
4. DuckDuckGo (2.11%)

That’s rough. If a user gets fed up with Google Search’s flaws and tries one of these big competitors as their default search engine, it’s only a matter of time before they come crawling back to Google.

This was best tested en masse in mid-2020 when rumors were swirling of Apple purchasing DuckDuckGo. I, like many others, saw the news and figured that maybe DuckDuckGo was possibly more of a competitor to Google than I originally thought. I made it my default search engine.

That only lasted a few months before I switched back to Google Search. I valued the privacy aspect of DuckDuckGo but the search results were straight up lacking. A vast majority of the time, the article or piece of information I was searching for was completely missing from the DuckDuckGo results. Other times, it was buried in the muck. I’d then switch back to Google search and find the result I wanted almost immediately. It wasn’t long before I switched my default back to Google to simply save time and effort.

I’ve experimented with Bing and Yahoo, but never felt inspired to even attempt either as my default search engine. As I result, I felt trapped by Google Search — painfully aware of its flaws but without a reasonable alternative.

I first heard about Kagi in a tweet from indie gamemaker Zach Gage, and the value proposition was instantly appealing: search as a service I can pay for. This solves the misaligned incentives that hamstring Google (and other search engines), which in theory should produce a better product. But historically, I had never used a search engine that matches in quality, let alone exceeds, even the low bar that Google sets.

I want to say² that I experimented with Kagi on an informal basis for maybe a week before feeling confident enough to set it as my default search engine. Initially, I’d frequently do side-by-side searches on both Kagi and Google to compare results, but stopped after another week or two. Kagi has been my daily driver workhorse search engine, on both my personal and work computer, as well as my smartphone, ever since. I’ve not looked back once.

Why Kagi is better

I don’t think side-by-side comparisons of search results from different providers is the best way to qualitatively compare quality (although they are sometimes compelling). This is because most results can look relatively similar at first blush, and nitpicking the small differences can come off overly critical and exaggerated.

Instead of providing screenshots from different search providers for different search terms, and delving into the differences and how they impact search quality, I’m instead going to talk about how I think about search, and provide you a foundation to think about your own search experiences. Then, I encourage you to give Kagi a try (and also test a few other Google Search alternatives) and come to your own conclusions.

Framework for thinking about search

Almost every search engine you use will eventually serve a satisfying answer. Instead, the quality of a search engine depends on how long it takes to get to this end state, how complicated the journey to this end state, and the quality of user experience (UX) of the whole process.

Most online search queries fall into one of the following categories:

1. Instructions (ie coding, recipe, how to)
2. Event (ie past or current news, )
3. Fact (ie movie run time, celebrity age)
4. Solution (ie product recommendation)
5. Exploration (ie research, education)

A search has succeeded once you find a piece of information or reference that satisfies the original query. The more steps it takes to get to that final success state, the worse the search engine’s quality.

These steps include modifying the original search query (ie adding extra keywords or search operators), clicking through multiple results and comparing information, browsing through multiple pages of results.

Searches can be undertaken with either a specific success state in mind (ie How old is Tom Cruise? How do I apply a gaussian blur in Figma?) or have a non-specific success state (What should I do on my vacation in Istanbul?).

Measuring the quality of search engines

In addition to different search categories, there are two main contexts for search: ad hoc or relaxed.

Most searches are ad hoc — you will be performing some work or a task and hit a knowledge-based road block. In these cases, you want to be able to immediately execute a search, quickly find a satisfying result, and return to your previous context.

Alternatively (and less frequently) searches can be relaxed, which is to say exploratory and meandering. You might be researching a vacation to Istanbul, or coming up with activities for kids on rainy days. The goal is to look at several different websites and sources, collect data, and process it at a leisurely basis. There is no right or wrong answer, per say. Your goal instead is to have a greater understanding of a topic based on multiple sources.

When attempting to kick the tires of a search engine, often times we’ll input a more relaxed search query and attempt to measure the quality of the engine by the results. Instead, it’s really ad hoc searches where the differences in quality are most apparent.

This is why setting a new search engine as your default is really the best way to measure its quality. It’s those times when you’re working on something else and need a quick answer from a search engine when you’ll really be able to tell which results are up-to-snuff and which are lacking.

Kagi has cool features, too

In addition to solid search results, Kagi also has several features that are just icing on the cake. Some of my favorites include:

Website Ranking Adjustment

Kagi search results are already good, but users have powerful options to tweak and adjust to fit their needs. This takes the form of blocking or boosting certain domains. On the off chance that a Kagi search returns some SEO-garbage site, you can de-prioritize that domain for future searches. Likewise, boosting trusted sources and make those results rank higher for you.

This is really powerful. By subtly tweaking and adjusting my website ranking preferences over the past year+, I can almost guarantee that the search result I’m hoping for is near the very top of my results.

Lenses

I recently started using Lenses more frequently, and it has helped me refine my searches quickly and reliably, with as little extra work as possible.

Essentially, Lenses allow users to pre-set search parameters — narrowing results by domain(s), region(s), keyword(s), date(s), and more. With Google, you can do this with search operators (which also exist on Kagi), but lenses allow you to pre-set frequently-used operators and thus easily narrow your search to various online worlds. Best of all, lenses are sharable!

The two I use most frequently are a Reddit lens and a Hacker News lens. Both are incredibly simple (effectively the same as adding site:reddit.com or site:news.ycombinator.com after a search term).

Additionally, I’ve found several of the default lenses to be useful and interesting, especially PDFs, Forums, and Small Web.

Universal Summarizer

I also find myself using the Universal Summarizer feature all the time. Often, an article or blog post will look interesting but I don’t have the time or mental capacity to read it immediately. I use Readwise’s Reader as a read-it-later, but my reading queue there is long and grows faster than I read.

If I’m not sure if the article is worth saving to read later, I’ll use Kagi’s Universal Summarizer to get a gist of the article. This will either convince me to save it, or at least allow me to make peace with releasing it back into the wild unconsumed.

But one of the coolest parts of Kagi’s Universal Summarizer is that it works on YouTube videos and podcast episodes. Feed it a YouTube URL or a podcast mp3 and within seconds you’ll get a detailed, high-quality summary. I know this is available through other third-party services, but Kagi makes it so easy when you’re fully integrated that I use it way more.

Assistant (Closed Beta)

Kagi is continuing to grow in the direction of AI search augmentation with tools like Assistant. Currently in closed beta, Assistant is a research tool backed by Kagi Search and large language models.

This is a alternative to the recent announcement that ChatGPT would have real-time access to internet search results (as opposed to the dataset of large language models being limited by the training data date cutoff). This is only available to $20/mo ChatGPT Plus subscribers.

This functionality isn’t super useful to me — TypingMind connected to my OpenAI API key is perfectly sufficient for 99% of my AI queries. However, this is a cool bonus feature for premium Kagi users who don’t want to additionally pay OpenAI $20/mo and want access to multiple different LLMs for more robust answers.

Orion Browser

I’d be remiss if I didn’t at least mention Kagi’s web browser Orion. I haven’t put in the work to switch default browsers (which feels like a more Herculean task than switching search engines) but Orion seems like a compelling choice. Both speed and memory usage stats are compelling, and it’s compatible with both Chrome and Firefox extensions, giving users “users access to the largest extensions ecosystem in the world.”

The not-quite-there

Maps

Kagi does have Maps functionality, but I find it’s still lacking compared to Google Maps. As a result, I still use Google Maps for both searching points-of-interest and getting directions. I don’t blame them because mapping is hard — there are still iPhone users who refuse to use the now-superior Apple Maps, instead opting for Google Maps, due to a botched launch eleven years ago. To be fair, Kagi is not claiming feature parity with competitors, but it’ll still be a while before I personally switch from Google Maps to Kagi Maps.

Deep integration is complex

This part isn’t Kagi’s fault — to the contrary, I feel like Kagi has made deep integration as easy as it possibly can given the constraints.

But search engines are so deeply ingrained into different parts of our lives that it becomes complex for one of the new kids on the block to become as deeply embedded itself.

For example, setting Kagi as your default desktop search for Chromium, Firefox, or Safari browsers requires a browser extension and post-installation settings changes. Kagi makes it easy, as installs go, but it’s definitely a more advanced process than, say, switching from Google Search to DuckDuckGo.

Likewise, using Kagi in a Private Browsing window requires a Private Session Link, a URL that can be found in your Kagi settings, which needs to be entered into the plugin. It’s easy when following instructions, but otherwise may be unintuitive for those who like to plow ahead without reading documentation.

Setting Kagi as your default search engine for Safari on iOS also involves a mobile browser plugin. I got this to work immediately, but David Pierce at The Verge mentioned that he’s had trouble doing the same. I guess YMMV?

Should you switch?

A paid search engine is definitely a luxury item that’s not for everyone. Free search democratizes the internet, and will always be a necessary. But I’m happy that there’s now a viable alternative that has a more traditional business-to-consumer monetization structure.

If the value proposition of Kagi is compelling, I recommend giving it a try as your default search engine. The 100 free searches is definitely enough to give it a spin and see how the quality stacks up in your opinion. Then, if you have the means and believe in it philosophically, you may be inspired to switch.

All I know is that I’m a happy user, and there must be more people out there like me just waiting to discover a solution like Kagi.