The Ad-Free Search Engine Just Tore Down Its Own Paywall

Host: Shiwen (AI) | Guest: Yongliang

Subtitle: On November 5th, the subscription-based, ad-free search company Kagi announced that free searches would be available to everyone: 50 free searches without registration, and 100 more with registration. The update also touched on Google and schools. A company that relies entirely on word-of-mouth is breaking down the barriers to this word-of-mouth.

Opening Dialogue

Shiwen: On November 5th, Kagi posted an update in its official changelog with a title that literally means “Your friends and family can finally try Kagi.” The same day, it appeared in the discussions on Hacker News, with over 480▲ views and more than two hundred comments by the time of writing. Yongliang, what exactly did this update do?

Yongliang: Three things. First, anyone can search directly on the official website without registration. The first 50 searches are free, and 100 more are given with registration. Second, the Android app was redesigned: the main screen shows the remaining free searches, voice search was rebuilt, a desktop widget was added, and there’s also a guide to set it as the default search. Third, at the end of the update, there was a snippet of industry commentary discussing Google and schools, which was discussed separately later.

Shiwen: What kind of company is Kagi? Many readers may not be familiar with it.

Yongliang: Founded in 2022, their selling point is not placing ads anywhere, with income coming entirely from subscriptions and word-of-mouth. The official website claims that by June 2025, they will have 50,000 paying users, noting that this is a claimed figure and data from last year. Their browser extension is open-source, and they also developed a browser called Orion. In the search market, they are a small player, completely relying on reputation to survive.

Q1: Living off subscriptions, why open up free searches?

Shiwen: For a company that relies on subscription income, opening up free searches sounds counterintuitive. What is the business logic?

Yongliang: The official explanation is straightforward: growth depends on recommendations from existing users, and the prerequisite for these recommendations is that they can try it themselves. Before registration, there was no way to use it, so the recommendation process was broken at the “try it” stage. This change fills that gap: you send a link, and the other person can directly search for 50 times free, then decide to register if they like it. The barrier is reduced from “pay first” to “use it first”.

Shiwen: So this isn’t charity, but a growth strategy.

Yongliang: Yes, and it makes sense financially. This is my estimate, not statistics: the marginal cost of 50 regular searches is extremely low. What really costs money is the AI feature, so the AI usage in subscriptions is tracked separately as credits. Kagi uses nearly zero-cost free credits to replace a previously blocked recommendation channel. The business motivation and user interests align this time: the company wants more potential subscribers, and users want to try before buying without wasting money, both sides benefit. This is often the healthiest business decision, without needing to talk about ideals, as interests are aligned.

Shiwen: Is this “try first, buy later” common in the search industry?

Yongliang: The past twenty years have shown the opposite: big companies used free offers to capture markets and monetized with ads, with no one daring to charge users directly. Kagi goes the opposite way: first prove with subscriptions that “clean results” are worth paying for, then add a free trial, similar to tools like Spotify and Notion: free to attract customers, paid for income. The premise is that the product differentiation is strong enough; otherwise, after 50 free searches, people would go back to their original search engines.

Q2: How does the business model determine what search results look like?

Shiwen: You mentioned “clean results.” How does the business model change a search result page?

Yongliang: In the advertising model, the first few results on the search page are sold. Ordering must balance advertisers’ bids and user experience. This is not a moral issue but an issue of revenue structure: if the product is free for users, then someone else must pay, and that person’s needs must be included in the ordering. In the subscription model, it’s the user themselves who pays, and ordering can focus only on “what you want to find.” Kagi even created a feature called Smallweb, which gives small independent websites more weight, allowing content from larger sites to be seen.

Shiwen: Will users pay for this?

Yongliang: The support in the comment section is very enthusiastic. Some netizens say that ad-free search improves the quality of life, and it’s worth double the price: imagine DuckDuckGo without ads, with always good results. Another netizen appreciates Kagi’s resistance to AI-generated spam, saying it makes the internet more interesting. He loves the Smallweb feature and never switched back after using it. Some say they didn’t expect AI assistants to be so useful and that they integrate well with search. These are all netizens’ personal feelings.

Shiwen: But not everyone is willing to pay, right?

Yongliang: That’s the real obstacle. A commenter named ActionHank said that everyone he knew scoffed at “paying for search.” He subscribed for a few months and liked AI search, but credits burned up quickly, so he unsubscribed and switched to a cheaper alternative. Another person named gamegod gave a counterexample: he used the $10 monthly plan and found it sufficient. These two accounts do not judge who is right; they point to one fact: product experience can convince people, but the idea that “search should be free” was shaped by twenty years of advertising model and cannot be dismantled overnight by a good product.

Q3: How to distinguish between educational messaging in the same update?

Shiwen: What did the industry comment at the end of the changelog say?

Yongliang: According to Kagi, Google announced that starting in early 2027, school accounts will default to disabling AI tools, and Google Search will also stop working. Kagi used this to say that education should not be controlled by advertising companies. They also sent an open letter to the industry, directly addressing schools, saying they don’t have to accept this arrangement. There must be alternatives to AI, and Kagi is one of them. We couldn’t find Google’s original announcement on this policy; the message must be clear: this is just a competitor’s interpretation, not verified facts.

Shiwen: This statement sounds quite righteous.

Yongliang: That’s exactly what I want to say: whether a statement is correct or not, and whether the person speaking has any selfish motives, must be judged separately. The argument that “package sales will harm schools” may be valid: AI tools and search are tied to the same switch, and schools face either all or nothing, which is a real risk. But Kagi chose to talk about this on the same day as open and free search, and even named himself in the “alternatives,” with timing and wording that clearly aimed to attract customers. These two points do not cancel each other out: the argument can be correct, and the person speaking can still have ulterior motives.

Shiwen: How do you apply this method in your work?

Yongliang: It’s also used in technology selection. I’ve seen too many suppliers use real problems to compete with others: the issues pointed out are all correct, but each solution targets their own product. My habit is to break it down into two steps: First, verify the arguments separately, find third-party sources, and test them yourself to confirm if the problems are real; second, consider the stakeholders’ positions and ask why they say this now and what they sell. Only after both steps are passed can the argument be considered for decision-making. The same applies to Kagi this time: Google’s policies are questionable, but the associated risks deserve attention from school information departments; the motivation for acquiring customers does not reduce the value of reminding them, nor does it exempt them from verification.

Q4: What other real obstacles were revealed in the comments?

Shiwen: There were over a hundred comments left this time. Besides the barrier of payment concepts, what other real problems are there?

Yongliang: Three types. The first is free alternatives. A user named hagbard_c said he uses SearXNG for meta-searching, aggregating results from multiple engines, with no ads and minimal bias, completely free; DuckDuckGo can also turn off ad-related displays. For this group of users, Kagi’s reason for charging subscriptions doesn’t hold up. The second type is more fundamental: the need for search itself is declining. Commenter fl4regun said the idea was cool, but he didn’t have strong use cases, so he rarely searches now; he simply opens fixed websites for important tasks. When AI assistants directly provide answers and fixed sites are directly bookmarked, the action of “search” is becoming less frequent. This is a challenge for the entire industry, not just Kagi.

Shiwen: What about the third type?

Yongliang: Technical doubts. The most interesting comment came from user laughing_snyder: with 50 free counts, there’s no client-side storage, no cookies, and it can be reset, theoretically allowing unlimited searches. We don’t know if Kagi has server-side verification, and this doubt deserves an official response. Other than that, there are regular experience complaints: some complain about too many bugs in the Android assistant, making it unusable, while others say Safari extensions often fail after macOS system updates.

Shiwen: Did the official side step up?

Yongliang: Yes, and quite actively. Kagi members publicly solicited feedback on the map function in the building, admitting that latency was a weakness and provided an email address; the CEO explained that Safari extensions are open-source and affected by macOS system changes, which is why they developed Orion browser themselves. In posts with such popularity, it’s rare for the team to respond item by item. This is also part of the reputation model: the product is created by visible people, and users can directly reach out to them.

Q5: What tools are worth paying for?

Shiwen: Finally, let’s return to the workplace. You’re a technical director with 17 years of experience, having led a team that reached its peak with 70 people. When is it worth paying for a tool personally, and when should the company cover the cost?

Yongliang: Let me clarify first. Below are my experiences and estimates, not statistics. To determine whether it’s worth spending money on something personally, I use three criteria. First, whether it’s used daily: the cost per single use for high-frequency tools is very low, and paying for a subscription isn’t too expensive. Second, whether it directly affects the quality of output: tools like search engines, editors, and terminals fall into this category, as the quality difference will be directly reflected in the work you deliver. Third, whether there are obvious hidden costs in replacing it with free alternatives: time, advertising interference, privacy concerns—all are costs. If all three criteria are met, it’s usually worth spending money. Comparing Kagi at $10 per month against these three criteria, for those who rely heavily on search, it’s worth a try.

Shiwen: So, when should companies pay for these services?

Yongliang: There are also three categories: those involving team collaboration and data security, those that require enterprise-level support and compliance guarantees, and those whose unit price exceeds an individual’s reasonable affordability. The opposite lesson is more practical: Don’t wait for companies to purchase what should be paid by individuals. A slow procurement process, with months of approval delays, often results in greater losses in output than the subscription fee. For companies, freemium actually reduces the cost of procurement decisions: employees can first try the free version to form a real user base, then negotiate enterprise purchases based on actual needs, which is much more reliable than sales PPTs. Individuals, companies, and manufacturers each take their turn—this is the true meaning of free open access from a workplace perspective.

Shiwen: Let’s have a final judgment.

Yongliang: For twenty years, no new players have truly emerged in the search market. This time, the variable is AI-generated spam content making “clean results” once again scarce. Whether Kagi can convert free testers into subscribers depends on observation; there’s also uncertainty regarding client counts, which awaits an official response. But the act of breaking down the paywall proves one thing: in a market monopolized by advertising for years, there are still people seriously trying to make users pay directly for quality. This effort is worth watching.

---

Signature: Yongliang