Anyone on an enforced-safesearch network can now go to kagi.com, turn SafeSearch off, and search for porn or anything else, which is just going to get kagi blocked completely on all those networks. If you add a SafeSearch-enforced alternative IP address, then Kagi can still be used.
[1] https://kagifeedback.org/d/3285-safe-search-dns-locking-for-...
Either way, ultimately the result of not offering the option will more likely be getting blocked entirely, than workers being freed from the shackles of SafeSearch.
Also blocking the search is pretty weird, what prevents me of typing the URL directly?
Not if the machine is locked down. Blocking all search engines is not a feasible approach, though, hence the Safe Search mechanism.
> Also blocking the search is pretty weird, what prevents me of typing the URL directly?
That URL can be blocked at the router level, but one can still see cached search engine images when safe search is off.
It's not just porn, it could be a way to prevent successful phishing through employees filling out a malicious form or whatever.
"Trust your workers" is bad infosec.
But blocking alternative search engines and AI engines goes a bit far IMO especially if they dont use training. Especially if you're just afraid they will see a nipple on a search engine. That's ridiculous.
Some companies just know their workers, i.e. if you have a bigger company you will have some idiots. We heard things like "you shouldn't be paying an escort with your company credit card" and the reason was because it was done, multiple times.
Can you expand upon this? Why is there an AI block?
The reason is that copilot resells other people's models (they're just a middle man) and there's an easy way to cut them out. So they blow up the dangers of other people's AI. Whereas if your data is in M365 they have it all anyway whether you use copilot or not. So they're trying to build a moat so they don't easily get cut out.
Not working with Kagi, but while they have RU contracts this will not be possible, can guarantee.
80% of the time the search results were perfectly fine, but that small amount of time would royally annoy me and my breaking point was this: https://i.redd.it/r4q6y3vl5vdh1.png
So I went back to Kagi and appreciate it a little more (they've also rolled back on Translate to some degree).
Come to think of it, one of my boxes is not logged into kagi, and is mostly ai nonsense.
I guess I’ll keep paying them. :-)
I’m not sure if this holds true nowadays. I’d say it git pretty bad, but now Gemini summary pretty much covers all of my use cases. And for any meaningful query, it’s straight to ChatGPT/Claude.
Is it though? How often are you spot checking sources?
My google queries are like different combos of weather, specific event questions, directly attributed to wikipedia data and etc. LLMs are way past the point of reliability when it comes to data extraction from static websites.
Two very modern counterexamples:
https://app.opencve.io/cve/CVE-2026-83844
https://www.sentinelone.com/vulnerability-database/cve-2026-...
The information they feed from is simple, clear, and static. And yet the both of them are hallucinating mitigations, compromise indicators, and other details of the vector that flatly don't exist.
The question GP posed wasn't bait. It's still a blatant problem in 2026. It's not just shitty cheap models, either. Just this morning I caught 6.1 Sol (via Codex) incorrectly answering a question about the codebase it was actively working in.
When I dismissed the popup, it produced much better results than screenshots I’ve seen from logged in browsers.
I guess that’s an improvement?
And I won't know which ones are the wrong ones, so I may as well trust it 0% of the time.
Most people I know use AI Overview now as their primary way of doing any sort of research or question-answering... and I urge them: If getting an LLM take is suitable for your purpose, at least ask free-tier ChatGPT, which is at least an order of magnitude less likely to vomit up false or confused information. Also, asking actual Gemini is another sane option that is far superior to AI Overview.
Most people are not asking google to build them a train signaling system but to look up some question on Wikipedia for which the answer does not matter that much
Think of the incentives and the path of absolute least resistance/thinking/energy use and you will arrive at the product people want.
Does truth matter in the post-truth world where disclaiming truthful news about you is as easy as calling it fake?
What a disgusting use of technology. Kicking a query to an agent, that's leveraging a model that has bulk of humanities information in it, so it can use a web browser to go looking for the thing you searched. WTF.
1. leverage the AI Overview model (with the IQ of a brain-damaged monkey),
2. run an ad auction and populate the first 6-10 items from that,
3. then do a perfunctory exercise of searching its spam-riddled index to round out a few results at the very bottom of the page.
Yeah, that guy asking ChatGPT is a monster.
Google intentionally serves bad results now to juice ad impressions (according to sworn testimony from their execs), so I assume DDG is significantly better these days.
- DDG
- Kagi
- Bing (10+ years ago)
- Yahoo (I'm getting old...)
I searched by typing lists of keywords into the search box in the past. Maybe you're one of those people that types sentences, or relies heavily on synonym matching? (Which finally works now, but only if you turn on LLM mode -- even then, Google shipped that months or years after kagi, and the reviews are scathing vs. the performance I get out of the Kagi AI search.)
[Citation Needed]
1. DDG
2. Google
3. Whichever LLM is closest at hand
4. This is information the world was not meant to know
(Also Marginalia for deep search)
As for Google, I typically only use its "Web" search so I skip all the junk Google adds to the `All` tab. With Firefox or Chrome, you can make a custom search engine with `&udm=14` appended, like `https://www.google.com/search?q=%s&udm=14` if you want it to be your default search engine.
The last time kagi failed me, I flipped on claude research mode. It found a microfiche archive, OCR’ed it, then started retro-stalking obscure 90’s celebrities (read old interviews, found coworkers, etc).
It took 15 minutes, but it produced an accurate 20-30 page report, then converted it to machine readable form for downstream use. I tried doing the same research myself a few years ago and failed to find anything at all.
As the kids say: Google’s cooked.
However, Google has shat the bed, itself, the room the bed is in, the neighbors and half of the city with the search. Since search no longer returns any semblance of relevant data, but ads and wrong AI summaries.
So for vast majority of use cases I find Kagi much more superior. And in 1 or 2 cases when I'm frustrated with the results I can append `g!` before the query to get the shitty Google results.
The only downside: Apple doesn't let you set a custom search engine to Kagi. And (a very unpleasant surprise for me) neither does Firefox. You need extensions.
(Does Chrome on mobile allow it? I use Edge on iOS which also only has a hardcoded list, so while I don't know for sure, but I assume not)
[1] no doubt i will be waiting until i am dead
Is the 20th most visited website in the world.
I suspect they are aware of their own leaderboards and use those in their ranking algorithm.
I understand that price also includes the AI stuff, but I’d rather use something like the ChatGPT app directly than go via Kagi for that.
Right now there’s $10 for unlimited search and AI stuff and $5 for limited search and AI stuff.
At what price point could Kagi offer unlimited search and no AI stuff?
Let me give you $1 or $2 per month (or $24/year) for my 10-ish search requests a week and you've got yourself a subscriber.
Maybe if it was a secure instance you could order for your organisation. Like a police force or consulting company, where you don’t want people eavesdropping on your traffic.
What is the value to society of having a search engine alternative that doesn't desperately mine and monetize every single thing about us they can?
It's not just a search engine subscription.
This could be a first step towards freeing us from advertising. An investment in an alternate vision that doesn't rely on rampant manipulation and consumerism.
The cost is, what, a few coffees per month? Can we afford to not support initiatives like this? We can steer away from a feudalist cyberpunk dystopia.
Yet, after all of that, I still don't want to pay $10/mo.
There were a few times where I would search for things and get some very, very weird results (e.g., when looking for GH repos, I'd find things at the top from places like gitmemory (I think that's what it was called)) and I thought that was very odd, but it hasn't really happened again in a few days so I don't know if it was just a bug or what.
I switched to Ecosia because I want to support European alternatives, but that's all. I tried Qwant for a while, and for some reason it was frustrating me (I assume that the results were not satisfying for me?). I wonder if Qwant is better for search in French? Not sure. Ecosia hasn't been frustrating though, and I cannot say that it's worse than my experience with Kagi. I just wish I could pay Ecosia instead of having ads (which my browser seems to block anyway).
Anyway, I always wonder why it is that a search engine (say Kagi, Ecosia) can "work well" and another one (in my case Qwant) can be "frustrating". If they all tap into Google/Bing and others, I would expect similar quality? I'd be curious if someone could explain!
Kagi uses a lot of different providers, including [Yandex](https://kagifeedback.org/d/5445-reconsider-yandex-integratio...).
I would recommend trying out [Uruky](https://uruky.com/) You can choose your own indexes to use. So if you want to support European companies you can try to use as much as possible European or open-source indexes and only when you can't find stuff to back to google's (they even have a private google index that they use called Serper). You can also configure fallback search engines like DDG, Brave...
But it seems Ecosia does really help around the world, at least looking at their reports certainly seems so. Even if it's not a huge impact, isn't it better than nothing? How does NOT using it help anyone?
Instead, just refusing an unnecessary work trip or something once a year helps a lot more than using ecosia.
There's a problem with "feelgood as a service" when the problem is so big it needs drastic measures. It gives people the impression that there is a quick fix. Whoop, I just changed my default browser, great, I've done something to solve climate change! But still eating 300g of meat every day, meh.
That's what I object with with ecosia. They talk of "saving the planet" on their website. But on a global scale it basically does nothing.
I genuinely don't care about their "green marketing". I use Ecosia because it is European, and it works for me as a search engine. As I said I tried Qwant but for some reason it was not working for me.
I do care a lot about ecology, but my choice of search engine is not really impacted by that. People who think that "planting trees" is helping the environment need to get informed about the problem.
Still better than using Google directly.
I agree with you, it won't change the world. But again I think it's mainly just marketing. They do what they claim and they do seem to care about actually making a difference and are somewhat consistent with their efforts so I wouldn't call that "greenwashing". I don't think their efforts are minimal but what do I know about planting trees and stuff. You can check their youtube channel :
If my agents are running internet search for me; why should I pay extra? My usage of Kagi has diminished since the raise of agents, so it should compensate. Maybe simply adding a soft cap to prevent abuse, but that's a bit lame to have to pay twice.
Right now, API usage is only available using a PAYG credit system, even if you've already got a subscription. But giving subscription users an allotment of free credits that covers light API usage would be nice - I could plug it into my agent of choice for casual chats or exploration, while still having to buy credits for more extreme or production usage.
I'm imagining rate limits that would prevent the user from making many more API calls than they would have made searches anyway. If you want unlimited 24/7 searches, you can pay for the credits directly, but if you just want to partially replace your existing search habits with e.g. interactive agent sessions, then that would be fine.
A more specific example would be how Anthropic allow certain Claude subscriptions to get free API credits, with some restrictions on the age of the account and with the number of credits being limited.
I think there is a workable solution here, although I can imagine that figuring out the details (how many credits to offer, what sort of limits, how to price the whole thing, etc) is complicated enough that Kagi have decided it's not worth the effort for the small number of users for whom this feature would be useful. And generally I'm very satisfied with Kagi, so I'm not complaining here or trying to suggest that this is something that they should be expected to offer. It would just be nice!
e.g. if you expect the average user to use 50% of the search quota / month through the UI, but 100% through the API, then you make the quota shared between the two but double the effect of usage from the API.
Especially if the API induces demand.
(Don't shoot the messenger, just trying to explain why it makes sense that they're not doing this)
[1] Meaning, they are more satisfied with an 'unlimited' or just way-more-than-you'll-ever-need option, even when it costs them more when averaged over even a few months. I'm not saying this makes you an irrational person -- I have the same illogical feelings myself.
This is why unlimited or at least a service stop instead of overage is highly preferable. It's the reason I only use prepaid mobile plans too so they can never overcharge me (and it's cheaper than a contract too)
When a real user is going through the trial period, those 50 searches get them familiar with the tool. When their usage runs out, they're forced to decide if they want to pay to continue or invest the work into replacing it.
AI agents typically don't care. They may sign up for a trial, use the trial, and then when the trial runs out they might get creative and create a new account, or silently switch to another service to avoid bothering the user.
Browsing some of the forums where people discuss their search setups is enlightening. The current favorite strategy is to have your AI set up access to all of the different search providers with free tiers and then rotate among them to avoid having to pay for anything.
Defeats the purpose of the free tier.
Interesting, is this for personal use or is it some business application? In any event I'd imagine the same can be done via agents using a browser.
> AI agents typically don't care. They may sign up for a trial, use the trial, and then when the trial runs out they might get creative and create a new account, or silently switch to another service to avoid bothering the user.
That is a crazy amount of compute just to conduct a search.
It's the people downloading open weight models and running OpenClaw (or this month's trend) who don't care how many tokens it takes because it only costs them electricity.
Real businesses aren't doing stuff like this.
It's the home users who use LLMs as a hobby. Some power users pay for Exa and other services. A lot of home LLM users will go to unbelievable lengths to not pay for anything that isn't hardware. Seeing what your agent can do is half the fun for them.
Set one of them against a goal and tell it to not bother you until it has the answer. They will do some weird things as they go.
But those are an offer and are used to advertise and get more users. Not to mention, as the "50/100" says, they're capped.
I don't get it, what other sense is there? Seems like 50 daily API queries would be lighter weight than 50 browser queries. And with an API you can require a key (aka subscriber).
> But those are an offer and are used to advertise and get more users.
"Try our search engine from your browser" / "Try our search engine from your agent setup" These both seem like user acquisition pitches. The real difference with programmatic access is that you'll likely get users hitting the cap and paying for more usage if they like the product.
> as the "50/100" says, they're capped.
... which is exactly what I was suggesting for API use.
Obvioulsy the TYPE and RETURNS of some cost.
>These both seem like user acquisition pitches. The real difference with programmatic access is that you'll likely get users hitting the cap and paying for more usage if they like the product.
50 searches from the web feel way more generous to a user than using the agent (most wont even hit them), hitting the 50 cap in like 10 minutes, and being negative about the whole experience.
>... which is exactly what I was suggesting for API use.
Which is exactly why it wouldn't work well for API use with such cap.
I refuse to believe there is a negligible margin decrease for agents using the regular search budget. Searching is cheap.
That’s a bad business decision in the same way it would be a bad business decision for Netflix to re-enable infinite account sharing.
A bit off-topic, but when I signed up for Kagi, I naturally became more aware of how many times I use search on the internet, since I didn't want to run into the limit for my tier.
I assumed I was a heavy search user, but now that I watch myself, I'm really surprised how little I actually search anymore. I thought it might be a hundred times a day. In reality, it's closer to four or five times a day.
I don't know if it's my changing habits; or the fact that much of the information on the internet has become siloed, so I know where to go without wandering around; or some other factor.
I was just really surprised.
Is it plausible that Kagi's financials require them to run a tighter ship than established players?
Conversely, is it plausible that they are so popular that they can afford to be deaf to the negative UX of the extra sub?
I would suggest that the answers to those, and a quick survey of the heavy costs agentic use has imposed on site operator over the past few year5, suggest that this is a reasonable compromise that keeps costs down for the average user and moves agentic users to the API-fee paradigm that is already rather common.
Weaker on actual search features though
I also see a lot less value from my Kagi subscription like the grandparent post than I did before agentic search. I’ve been wondering if I should just drop it in favor of DDG; allowing me to use some search via api would change that for me.
You're paying ten bucks a month for access to Kagi the web search, not Kagi the search API. If you want them to provide that service and scale it accordingly it makes sense for that to be a separate feature with a separate price.
The big difference is that API usage can be scaled up way quicker than web page hits and scraping can be, which is an understandable concern if you're trying to price your service for human usage and not for bots hammering away as fast as they can. But judicious use of rate limits could help here - have a subscribers API tier that allows only a relatively low volume of calls and do standard API pricing for all requests beyond that.
That said, I'm sure there are complexities here that I'm missing that make this easier to say than to do.
I suspect Kagi would actually make more money by including a small recurring API allowance with existing subscriptions. The value ain't really in giving away searches, it's getting people to integrate Kagi into their existing agentic workflows.
Once someone starts relying on Kagi for their agents, they're much more likely to exceed that allowance and pay for additional usage. That seems like a better long-term revenue strategy than trying to convince people to pay separately for API access before they've even built anything around it.
Brave and Perplexity seem to understand this. A small included allowance is a much more effective customer acquisition and retention tool than keeping API usage entirely separate.
Hopefully we can win you over soon enough <3
- dylan (maps team)
1. Transit. I'm in NYC, and Google Maps does a great job of showing the MTA, trains and busses. Every route, all the stations, which trains stop there and when, even where the entrances are. Kagi Maps has no train routes or layers. There's stations, most of them at least, but it's basically the Wikipedia entry for that station. No up-to-date info or routes.
2. Mobile apps. At least half of the time, my maps use is an Android intent for address or directions, which only an app can answer. And the only really good way to get your location and see what's around you.
I don't really notice latency at all personally, but fixing those things might bring my Kagi Maps usage from 0% to like 50%+.
I know, it's probably a pain in the butt to integrate with the transit systems of hundreds of cities for up-to-date info, but OSM does at least have the stations and routes. Not much point trying to beat the official MTA app in functionality, but something is better than nothing.
It's also weirdly hard to get to. Why isn't it an option in the main list of Kagi things? To get there from the Kagi homepage, I have to click on search options, then maps.
But Google Maps has been getting worse over time for me, and I want something better. GM results contain items that are completely irrelevant to what I search, selecting "4.5+ Stars" filter returns places rated only 3 stars and hides valid 4.5+ star results, "open now" includes places that the results themselves say are closed, the list goes on...
I just went through my recent !gm history, compared it to Kagi, and I'm impressed with the results and that things like ratings are there, but do have a few notes.
- Filters. Sorting is great, but filters would be more ideal. Example use-case for me you can probably guess from my gm gripes. Finding restaurants with filters for ratings, cuisine, "open now" and/or open hours, and distance. The ratings and "open now" filters would be enough for me to switch.
- Searching for something, then panning away or zooming feels like it's missing a "search this area" button, since the results don't automatically adapt.
- Map initializes in the correct general area. Clicking the crosshairs location button shows a "Couldn't find your location" error, even after allowing location in the site permissions and refreshing a bunch. (reproduced in Brave and Safari. Chrome doesn't seem to have this issue, same permissions prompts for browser/OS popped up for all three)
- Opening the "Control Center" right sidebar and scrolling randomly produces artifacts, like it's flashing by disappearing/reappearing. (reproduced in Chrome and Brave, Safari seems fine)
1. Navigation makes pointless detours.
2. The "rotate the map back to north-up" button doesn't work.
3. "You're looking at Barcelona and asked for restaurants? I know a few, in Tokyo"
4. Half of navigation window is useless UI elements I cannot hide.
I've been trying mapy.cz for a while because they were fantastic for tourism purposes, but they're locking more and more features behind a paywall and I'm not sure I want to pay for that. The killer features of Google Maps are the POIs, reviews, photos, Street View, and public transport planning. I cannot really switch to another map app without having those.
I guess the biggest problem is and will always be the quality of data. The reason I can't use CoMaps much is because OSM just doesn't have the same accuracy and recency of data on businesses and POIs in my area as compared to Google/Apple.
1) navigation
2) give me more info about this business that I tapped on
3) reviews; find related businesses, yelp style
I cannot figure out how to do any of those things with your UI.
Ideally, for (1), it’d let me choose to open in organic maps (reliably works at all) or apple maps (better when it works, which is rare offline, even with offline maps turned on) on my phone.
This might he fantasy land, but if it could send the location to BMW’s app, then that’d be awesome too. (The bmw app can send the location on to my car.)
One other thing: Lot’s of people are asking for an app. “Add to homescreen” works well on first glance, at least in iOS. The big thing is integration with the other apps. (Something like avenza would be cool for parks.)
> For example, searching Kagi for !w Monty Python will search Wikipedia directly for "Monty Python" and take you to that Wikipedia search result. [1]
If you're on a plan with limited searches per month, it does not count against your quota.
DuckDuckGo has bangs as well.
This said, it's a good offer. You can switch between many models quickly. They are rated by quality, speed and price, and it's shown with orange dollar symbols for price and bullets for speed and quality. Example:
Anthropic Claude Opus 5.5
Anthropic's flagship model with always-on adaptive
thinking ($$$$$). Best for deep reasoning, agentic
coding, and long-horizon work.
Cost $$$$$
Quality ●●●●●
Speed ●●For those that are looking for it, you can track your current month's billing/ai-usage here: https://kagi.com/settings/billing
The last time I used it heavily was a few months ago so it might have gotten better, and I assume the more expensive tier where you can choose which models it uses might be a bit better depending on which model you choose. I much prefer just pointing my own local agent harness at a search api at this point though.
But since I'm on the Ultimate plan, for stuff that matters I usually use Fable with a custom assistant (prompt), and sometimes the stock "Research" assistant. Both work really great for me. I use the "Research" assistant for more technical questions, it tends to dig up information from forums and cross-reference it with the source-code file it digs up from github and quote the RFC that it has directly fetched too.
Though I usually prefer my custom assistant prompt, which Kagi seems to combine with their general prompt in the background. This works really, freaking great most of the time. My custom prompt is an elaborate version of "answer the question, but then try to second-guess what the user is doing and also provide a different perspective that could be more useful than a direct answer". This has derailed my narrow train of thoughts a few times and turned me around into the right direction.
That said, I haven't played around with my own local agents for search yet to compare (in part because Kagi does a very good job at this), and this is something I want to try, in a "own your tools" sense. But the first thing I'll make sure that it uses the Kagi API for its search.
[Cloudflare Turnstile] Error: 600010.
600* Generic challenge failure Bot behavior detected.