Kagi vs Google Search?

I was intrigued by the most recent episode of MPU, Exploring Kagi with CEO Vladimir Prelovac. I’m thinking about giving Kagi a trial run to see what I think. At $25/month for the Ultimate plan, I’m curious whether it will be worth the cost. I haven’t subscribed yet and have only run a few searches, but here’s a comparison so far: free Kagi (left) vs. a standard Google search (right).

I want to emphasize that this is using Kagi’s free search and is just an early comparison. With that caveat, while the Kagi version is cleaner, I don’t see a substantial difference in the results. (I’m not considering privacy here—only the search results.)

Is anyone using Kagi? What are your initial impressions?

2 Likes

As I wrote on my blog, I decided that it was not worth it for me (after paying for a month). Although the results were better than Google’s (certainly in not having the results determined by sponsorship and whatever opaque algorithms Google was using) and there were sites surfaced that were interesting (especially if I set Kagi to focus on academic material or the small web) it was not so much better that I wanted to pay for it. Worse was the feeling that Kagi had its own algorithms and approaches, some of which were a little dubious, and that they were no more transparent than Google in what these were.

4 Likes

I signed up and a still using it. I find the results good, I had been using DuckDuckGo until recently but like the idea of paying for a service rather than being the product.

It’s a bit clunkier to use on mobile as it’s not a supported search engine in Safari but there’s a guided workaround.

1 Like

I think it’s important to note that the professional plan has unlimited searches for $10/month (the only difference in the ultimate is access to the latest llms from OpenAI in Kagi Assistant)

There’s also no difference between any of the plans search responses.

In general, my experience has been that Kagi is on par or better than Google (especially once you down rank spammy sites) but I imagine that depends on what you’re searching for.

6 Likes

I’d never heard of Kagi before. And tbh I think this episode of MPU was merely “sponsored content”.

2 Likes

I think it’s David getting to talk to the developer of a tool he really enjoys using. With all the interaction I’ve had with David and his content over the years, I can’t imagine him doing a “sponsored” show.

7 Likes

I concur, David and Stephen have consistently reflected only the highest level of professional integrity. I thought it was an informative show on something I knew nothing about.

4 Likes

There are listeners (like me) who are fans of Kagi. Podcast hosts I like + founder I like = good episode.

Daring Fireball also had an interview with him late last year. Daring Fireball: The Talk Show: 'A Professional Internet User', With Kagi Founder and CEO Vlad Prelovac

Kagi vs. Google, in my experience, comes down to how much customization you do of Kagi, but the customization is easy to do.

2 Likes

I’ve read good things about Kagi, but I relied on Google search almost every day for nearly 20 years before I retired. I can normally find what I need on the first or second attempt so I haven’t even tried using AI for search.

If it “ain’t” broke . . .

1 Like

I think the quality now is different from 10 years ago :upside_down_face:

But of course search is definitely a personal thing. I wouldn’t claim universally that Kagi is better (in fact, at least from this old blog post it’s not even the best A look at search engines with their own indexes - Seirdy ) but in my experience it’s been the best of the ones I’ve tried

2 Likes

Google keeps making changes and not all are for the better. But many of the tricks I learned over the years still work.

1 Like

Thought you might be interested in the search results from Lumo the AI assistant by Proton

2 Likes

Thanks, I’ve never heard of Lumo. I wonder if the AI “revolution” will be able to significantly disrupt internet search and Google specifically. This will be interesting to watch unfold.

1 Like

I’m starting to see more articles about how the browser may become the “killer app” of AI. And that, some say, could disrupt more than search.

“OpenAI and Perplexity are taking a similar route, but with AI answers and assistance as the main feature,” he said. “For people already using these tools to search the web, manage tasks, or support their workflow, that kind of integration could be a big draw.”

2 Likes

I just started fiddling with AI last week, not only
how to use the tools, but also learning how to self host LLMs.

I do see a future where web search is done from an AI prompt (chatbot or no) instead of entering search terms in a search engine’s Web site and parsing the results.

this “clickless search” paradigm could really hurt ad revenue unless sponsored and unsponsored ads are integrated into the AI results.

It will be interesting to see how this plays out!

Back on topic, I skipped the Kagi episode as I’m happy with my current tools. If I change anything, I would switch to an AI prompt.

1 Like

For other benefits I already value, I have a $20/month paid ChatGPT account. I don’t see myself paying for a web browser/search tool too.

Power users might value some tools to pay for them, but I’ve seen a lot of pushback on subscriptions and monthly fees from a wide swath of users here and podcasters in general.

“Share of wallet” for monthly fees is a huge issue and I admit to pruning my list of paid apps for similar reasons.

I do a lot of my web searches now using ChatGPT instead of any browser.

I don’t know about absolute accuracy or less ad-supported bias, but the clean, crisp output from my queries in ChatGPT has been just as useful and much less ensh*tified.

I wish Kagi luck, but they could be “too little, too late” as all the huge AI companies launch their own browsers and aim squarely at the Internet search market.

2 Likes

I’ve have been using (and paying for) Kagi for a year and a half and a while ago I have reconsidered whether it was worth it. In particular, I was considering switching to a paid plan of Perplexity instead because I am really impressed by the results it gives me. I also like that subscribing to a Perplexity plan also gives you credits for using their API (something that I hope will become a standard in various AI providers because sometimes using the API is just the better or even the only way to achieve your goal, especially in automations).

Without going through the details of my reasoning, let me say that one of the main reasons for signing up for Kagi was to let my then 10-year-old daughter start searching the net without teaching her that Google is the only or even best way of doing that. I also liked the “Quick answer” feature as a way of teaching her that the AI’s answer is based on the webpages listed below (which can contain mistakes) rather than from a mysterious machine that knows everything. - That was before Google added AI summaries to its search results. When I now sometimes use Google for whatever reason, I tend to think that the Kagi Quick answers are better, but that may well be confirmation bias. I haven’t done any systematic comparison.

One thing that has annoyed me with Kagi from the start is that it lacks a product search like Google (and it turned out that I quite often search for products/ prices). In the beginning, it was difficult switching to Google on Safari but it works better now.

The main reason why I stopped thinking about unsubscribing from Kagi was that I discovered Kagi Translate, which I now use a lot, especially the proofreading feature. I created an Alfred Workflow that makes checking/translating any highlighted text a breeze. No copying and pasting and typing a prompt.

1 Like

Intriguing… According to Perplexity, Kagi is better than DeepL for translation. Thanks for sharing!

I gave it a sort of 5 search trial after reading that other thread, but it didn’t really stand out. Given the other positive comments in the thread, I gave it another whirl on my phone and used up the free 100 searches. But I’m none the wiser about why I’d pay for this from a performance perspective. The results were no better than Start Page and much worse than Google.

That said, traditional searches make up an increasingly small proportion of my work anyway, I use AI searches maybe 3/4 of the time. So I’m probably not the target audience.

1 Like

I agree. And I have been someone going around for years telling people who were using Google translate that DeepL is the way to go. But I have been increasingly disappointed with DeepL, especially when it came to translating individual words where it more than once gave me outrightly wrong translations (I think it works best for sentences of paragraph or even longer documents, which I haven’t tried yet). In addition to that, DeepL’s Mac app has been annoying me with extremely frequent updates which it, for some reason, doesn’t install automatically but bugs the user to confirm. Maybe it’s a setting, but I’ve never had an app that has annoyed me so much by just being there. LOL.

I think you might be misunderstanding Kagi. I’d say Kagi brings together the best of search and AI in a transparent way (this is no longer their unique selling point, but for some time it was). It sits right in the middle. You can do traditional searches. Or you can search and let the AI summarize the results for you (by simpky ending your query with a question mark). That is the Quick Answer feature. And if that isn’t enough, i.e. if you want to follow up from the Quick answer, you can continue the conversation with Kagi Assistant, which lets you choose among a variety of models. So you can seamlessly move from traditional search to an AI chat.

I’m not sure what the limitations for the assistant are for the free plan and if it can be properly tested without paying, but I’d say that “doing AI searches most of the time” is not a reason to dismiss Kagi.

1 Like