Provide a plan without AI functionality

Pleonasm

atom I have no illusion that Kagi will be able to outcompete the big guys on the AI stuff...

However, does Kagi believe that it can "outcompete the big guys on the AI stuff," based on the unverified assumption that the Kagi Assistant is superior to the solutions provided by the "big guys" because it is powered by Kagi Search?

atom People come here for the search. People are paying for the search.... Please, pretty please, focus on the search.

It mystifies me why Kagi has apparently lost sight of this fundamental marketplace reality, and apparently insists on prioritizing so many other ancillary initiatives above a focus on its core search service. But, doing so is Kagi's decision, and Kagi does not need to change its priorities. As W. Edwards Deming noted, "It is not necessary to change. Survival is not mandatory."

atom

ax7899 There are likely quite a few people, heavy AI users, prone to abusing the system. The spikes could very well be occuring just before Kagi plugs a hole in their enforcement of the fair-use policy.

Pleonasm An important observation, at least for me, is that Kagi does not need to compete with Google on pricing for search. Simply because Google is free (well, people here are of course aware of how it is paid for). This is critical, since Kagi would obviously not be able to compete on pricing with someone who sets the price (i.e. for the API access). However, the same does not hold true for the frontier AI, there Kagi is trying to compete with players who hold the keys to everything, including the pricing. The moment they stop subsidizing access to their models and decide to jack up prices, Kagi's AI is cheeper for us than search might start punching a hole into their finances.

atom

I did not think it would happen so fast, but the blog post about the translate makes it clear Kagi underestimated the token costs for the AI they use to provide the service or these costs have recently shot up. It would be absolutely understandable if this was the case only because of a skyrocketing number of free users (i.e. providing the service for does not make sense). However, their wording makes it clear this is not the case:

That might mean different limits for free use, account-based access, changes to what is included in each plan, or a different pricing model. We are still monitoring costs, so we may also need to adjust limits or pricing for paid users in the future.

The translate actually makes a lot of sense to me as an accompaniment to the core search service. I would much rather it is preserved in the pro plan than the assistant.

All in all, the whole situation does clearly reveal the folly of pushing for the AI. Kagi is stretching itself thin financially because it can't predict how the costs for the multitude of AI services they are trying to provide develop. In the end, we might really end up with a plan without AI functionality and hopefully it will be the pro plan at the current price. If they instead end up increasing the price of unlimited search plan trying to accomodate the AI costs, I think it will spell the doom of Kagi. And that is certainly something I do not want.

Pleonasm

atom ...Kagi underestimated the token costs for the AI they use to provide the service or these costs have recently shot up.

"GPT-5 launched in August 2025 at $1.25 per million input tokens. Eight months later, GPT-5.5 launched at $5, a 4× increase on the flagship line. Claude Opus 4.7 held its sticker price at $5/$25 but introduced a new tokenizer that uses up to 35% more tokens for the same text, an effective price hike Anthropic disclosed in its own pricing docs. Gemini 3.1 Pro is 60% more expensive on input than Gemini 2.5 Pro."
"The same shift is showing up at the consumer-product layer. GitHub just announced new Copilot Pro multipliers taking effect June 2026: Claude Sonnet 4.5 goes from 1× to 6×, Sonnet 4.6 from 1× to 9×, GPT-5.4 from 1× to 6×, Gemini 3 Pro from 1× to 6×, Claude Opus 4.6 from 3× to 27×. These aren't new flagship releases. These are the same models people are using today, repriced because the underlying economics no longer work at the old rate. When platforms repackage existing capability at 3-9× the cost, it suggests the cheap-access window we'd grown used to may be closing across the stack, not just at the frontier."

Source: LLM progress is slowing and getting more expensive

Pleonasm

Assisting If it comes right down to it, I'd pay a further premium to have it {AI} removed from the UI entirely, simply so I never have to think about it again. That does have a real value proposition for me, even if on the surface is seems like "paying more to lose a feature".

Interestingly, Brave has recently adopted this "pay more, get less" strategy with the introduction of the Brave Origin premium experience, providing a minimalist version of the browser with the 'bloat' removed for a one-time payment of USD$60.

Would Kagi users pay more to suppress AI and the rest of the 'Kagi ecosystem'? The new uruky search service suggests an answer: "We're building a private search engine, not an ecosystem. There are already many privacy-focused alternatives for other services controlled by big tech."

Anonymous45

atom I did not think it would happen so fast, but the blog post about the translate makes it clear Kagi underestimated the token costs for the AI they use to provide the service or these costs have recently shot up.

atom All in all, the whole situation does clearly reveal the folly of pushing for the AI.

The writing has been on the wall for quite some time. Several users, including me, tried to warn why AI is a bad idea both ethically, practically and economically, and now that the subsidizing party is showing clear signs of duress, everyone will learn the actual costs of providing and running LLM models.
After all the shilling, the insistence that users that don't like it should just not use it, and the endless rationalization that most users value these features and it was a necessary for Kagi as a company, I admit I'll find it personally hilarious if things get so expensive that either pricing for everyone shoots up or we end up with a cheaper AI-less plan despite all the reasons why that wouldn't be feasible.
Once you need to pay from 8 to 14 dollars per token, we'll find out just how "essential" using AI really is.

I left Kagi because it became increasingly clear they are an AI startup masquerading as a search engine alternative to big tech. Kagi's whole team is just clearly much more interested and focused in developing and experimenting with AI. They might just have to remember what brought most customers to them in the first place. We'll see how it goes.

Pleonasm

Anonymous45 Kagi's whole team is just clearly much more interested and focused in developing and experimenting with AI. They might just have to remember what brought most customers to them in the first place.

My own perspective has evolved, and I am in agreement that Kagi seems determined to compete in the marketplace as an AI company while neglecting its core search product.

The decision seems counterfactual, since Kagi appears to have no unique value proposition with respect to AI. While the Kagi Assistant is generally "good enough," the quality of its output consistently fails to match (and never surpasses) that of other top-tier tools (e.g, ChatGPT 5.5 Thinking), in my experience. For this reason, it is difficult to imagine that consumers will become customers of Kagi because of its AI tools.

Pleonasm

Anonymous45 Once you need to pay from 8 to 14 dollars per token, we'll find out just how "essential" using AI really is.

The issue of AI cost increases now has a name used internally at Microsoft: "Tokenpocalypse."

"This whole ecosystem is heavily, heavily subsidized by investor money. And so stuff that seems like it has no cost is, in fact, incredibly expensive. And now we’re going to get to a point where more of that cost is going to get passed on to the end consumer, to the customer."

Source: Is this the dawn of the Tokenpocalypse?

Anonymous45

Pleonasm Yup, it's a matter of time before OpenAI and Anthropic switch to token based billing, which will affect the whole market whether we like it or not. And the problem is that LLMs are fundamentally unreliable both in their output and the costs to generate said output. Any one action prompted to the model could burn as much or as few tokens in seemingly random fashion, no one can really predict how many tokens will be used and therefore financial costs are also unpredictable.

WatersLethe

I was excited to hear about Kagi, and was planning on signing up for me and my family, but I cannot support Kagi if all currently offered plans include AI features. From all the official responses above it seems clear that they don't care about anyone's moral stances about AI and aren't watching the growing popular anti-AI sentiment. It feels like a company that secretly likes AI more than they're letting on, has resentment to those who are opposed to the AI features, and can't read a room.

I'll subscribe to Uruky until I see a subscription option that clearly shows they understand why I've left Google.

beefz0rz

I really wanted a subscription to Kagi, but now that I have looked at the plans (all of which force you to also subscribe to their slopbot) I am on the fence about the whole thing.

I have stopped using every other company that has forced AI into their product (with the exception of my doctor who, after receiving torrential negative feedback, got rid of the phone-slop bot and re-implemented human receptionists).

Even if it is "opt-in" as they say, it feels like because of the way the subscriptions are structured, some of that money will inevitably be going to the AI perverts.

A customer service representative ensured me that the slop mongers would not receive any of my money if I didn't use their features, but that seems unlikely to me due to the fact that the two subscription plans are intertwined, you can't buy one without the other, which means I would be indirectly subsidizing these evil corporations anyway.

They're using their slop tech to steal from everyone, and are attempting to monopolize knowledge (and 'truth'). They're setting up a surveillance state to violate our privacy and strip us of our "rights". They build a CSAM generation machine, and refuse to implement parameters that would stop the machine from generating more CSAM.

I really want to try this search engine, but I do not want to subsidize the most evil people in the world in order to access it.

I can guarantee that if you implemented a "Zero AI" subscription, it would rapidly become your most popular subscription.

Until then, I will instead subscribe to the Uruky non-AI search mentioned by another user.

Please end your relationship with the pedo-slop-fascists.

beefz0rz

Vlad

I just wanted to concur with the user you are responding to. I agree with some of what you're saying as well, but not all of it.

I would also pay more for a zero-AI plan, but I would be resentful about having to do so.

A functional "remove all AI features" button would work as an alternative, with the downside being that you would be left with no metrics by which to judge how much people want to use your AI services (which is the current state of things, if I'm understanding correctly).

I do understand that there are certain accessibility functions that generative AI can be useful for, and I'm not against that; without going into personal details, I also have lived experience with disabilities, and I sympathize with that. However, part of the problem with "AI" is how the term is used as a blanket categorization for a bunch of different technologies. This is intentionally done on the part of the LLM corporations of course, in order to conflate their product with products that are actually useful (such as translation engines, or radiology imaging recognition, or speech to text services).

The way you describe the Kagi news summaries sounds interesting, but I'd have trouble taking anything it says seriously due to my previous experience with LLMs demonstrating that they are often wildly incorrect (for example, I asked a coder friend of mine about an issue I was having with my Linux PC. He queried Claude with my issue and linked me to the output. Claude [the amazing coding chatbot] output a terminal command that was half Unix-syntax and half Windows-syntax; I could recognize it as such, and I'm not even a coder, and this is the model that's explicitly supposed to be "amazing" at this one specific task). I do not want the possibility of accidentally querying one of these LLMs and absorbing it's hallucinations and fabrications. Accidentally executing a slop terminal command could be potentially detrimental to my system.

I concur as well that having to avoid pushing buttons that are strewn all about the interface doesn't really count as "opt-in" in my book.

Regarding the Kagi news summary bot being superior to news websites: Do people actually still visit news sites? They're so flagrantly homogeneous corporate mouthpieces. I subscribe to newsletters from several trustworthy independent journalists, and I doubt an LLM could offer a better experience than that.

Outside of some hyper specific accessibility use cases, I don't really see any benefit to this technology whatsoever.

Finally, I can tell from your engagement already that you are indeed different from "every company", that's why I'm even taking the time to type all of this out. You clearly have the capacity for introspection, and you are in here listening to constructive criticism in a way that most companies would not.

That is also why I have confidence that you can figure out a way for those of us who don't want the slop in our lives to be able to easily use your search without having to think about whether or not a given action will invoke the lie machine.

carl

Having customers complain when you're delivering a bad product is one thing, and quite expected.

Having customers complain when you're delivering a good product is another thing, not fun but part of life.

Having non-customers and non-users endlessly complain about your product which they didn't buy and don't want to buy, while your paying customers are happy - well that requires some patience.

Anonymous45

carl You're just an endlessly goal post moving, passive agressive poster, aren't you?
I'm an ex-costumer that left specifically because of this. Others here asking for no AI are paying customers as well. Potential customers are telling the company what it would take to become customers, you know, the things companies usually want. What's being asked isn't even unreasonable.

Then again, you're the guy that was claiming a lot of quite frankly gross assumptions about Brazillian people's economic situation back in the regional pricing thread, so I guess I shouldn't expect any different. The only thing that takes patience here is tolerating your taunts and provocations. If you're a happy customer, just be on your way.

carl

Anonymous45 gross assumptions about Brazillian people's economic situation

It's very off-topic, but go to Brazil and you'll see for yourself that there are very rich and very poor people there. And all kinds in between that. I remember that thread and that people were talking about currency exchange rates, which have nothing to do with somebody's economic situation. And talking about it like all Brazilians have the same or even comparable economic circumstances is far from reality.

Back on topic: It's been a while now that Kagi has continued to develop their AI offerings. It's clear that it is a part of what they are and do as a company. They clearly believe that offering AI features is what their customers and potential customers want. Probably they can get some new subscribers by bowing to anti-AI sentiments, but they could lose a whole lot more subscribers at the same time. They're a business, so it has to be along these lines they make their decisions.

But I guess they brought it upon themselves by promoting Kagi as an anti-google. This attracts users and customers who buy your product because they dislike the other guy, not because they love your product. I wish Kagi marketing would instead only focus on the benefits of their own products instead of complaining about the state of web search. "Reclaim the web, restore your privacy" is not a good slogan IMO.

Anyway, there's Uruky now for everybody who didn't like Kagi. That's why I don't understand the continuing complaints.

Anonymous45

carl sigh
Carl, I am Brazilian. Please, just quit it, I guarantee whatever you have to say next will only make it worse.

carl

Anonymous45 Then you should know, and not have to fake outrage. Inequality in Brazil is for Brazilians to solve, not for Kagi.

What did your lunch cost today? Is it not fair that one month of Kagi costs the same as eating out for lunch one or two times?

nobodywasishere

carl Please keep the discussion on topic.

Pleonasm

carl They {Kagi} clearly believe that offering AI features is what their customers and potential customers want.

It may not be the case that potential customers of Kagi place value on the AI services offered by the company. Examining the Kagi Stats, the rate of customer growth appears unchanged since the recent improvements in Kagi Assistant. This may be evidence that the marketplace is not eager to subscribe to Kagi because of its AI services.

P.S.: Of course, Kagi has a more thorough understanding of customer acquisition trends, and I encourage a Kagi team member to correct this analysis if it is defective.

carl

Pleonasm Then Kagi's non-AI alternative should be a block-buster success if that's the case. Maybe that's the case, time will tell.

Personally I think the key to Kagi's success lies only in search quality and ease of use, AI or not. That's why I've been on my hobby horse and on their case for years now about having an iOS app instead of a Safari hack. But I'm not ranting about pedos and fascists etc to make my case, as the commenter I originally replied to did. I thought that was a little too strong.

« Previous Page