Neeva had ex-Google executives, $77.5M in VC funding, and a full marketing team. It ran for four years and shut down its consumer product in June 2023. Kagi has one founder, $5.5M total raised, and 50 people. It reached profitability in 2024 and now has 66,000 paying subscribers.
Same thesis. Same market. The gap wasn't product quality. It was what happens when your investors need a billion-dollar exit and your cost to acquire a customer never gets low enough.
Founded by Vladimir Prelovac (CEO), a Yugoslav-born engineer who previously sold ManageWP to GoDaddy, served two years as VP Product, and quit to build the search engine he couldn't buy for his daughter.
This Week's Breakdown: Kagi
What they do: Kagi builds a subscription-funded, ad-free search engine and browser. You pay $10/month and get search results optimized for you, not for advertisers. No tracking. No sponsored results. No behavioral profiles. They also ship Orion, a zero-telemetry WebKit browser, and a multi-LLM AI assistant that comes bundled with the subscription.
The numbers: 66,000+ paying subscribers. ~$6-7M ARR. ~845,000 searches/day. Profitable since 2024. $5.5M raised total ($3M from founder, $2.5M from ~93 community angel investors who were already paying users). Team of 50. Zero paid marketing spend. Ever.
How It Actually Works
You type a query into Kagi and get results that no one paid to appear. Behind the interface, Kagi is a metasearch engine. It pulls from Mojeek, Brave Search, its own Teclis and TinyGem indexes for personal blogs and small-web content, plus licensed data from Yandex and specialized providers like Wolfram Alpha. The results pass through Kagi's ranking layer before you see them.
That ranking layer is what makes Kagi structurally different. Every search result has a small shield icon. Click it and you can permanently boost that domain (so it ranks higher in every future search), block it entirely, or pin it to the top of your results. Your preferences persist across sessions. Over time, Kagi's results get sharper for you specifically, not because it tracked your behavior to sell to an advertiser, but because you trained it on your own standards.
The hard problem Kagi solved is alignment. Google's incentive is to keep you clicking. That is not the same as answering your question well. When Kagi is paid directly by the person searching, those two things become identical.
The Key Bets
| Decision | What they chose | What everyone else does |
|---|---|---|
| Business model | Subscription-only: users pay $5-$25/month, zero ads | Ad-funded "free" search that monetizes attention and behavioral data |
| Personalization | Users permanently control their own result ranking — boost, block, or pin any domain | One algorithm tuned for the median user and optimized for advertiser click rates |
| Index strategy | Curate quality over scale: own "small web" index of personal blogs alongside aggregated sources | Maximize crawl coverage and surface content that drives ad clicks |
The Real Story
Vladimir Prelovac grew up in Yugoslavia, a country that no longer exists. He wrote his first programs on a Commodore 64 and spent the next two decades building bootstrapped software. ManageWP, a WordPress management tool, grew to tens of thousands of customers before GoDaddy acquired it in 2016. He moved to the Bay Area, took the VP Product role, and stayed two years before deciding corporate life wasn't for him.
In 2018, he had what he calls a "Matrix moment." He was searching for GitLab on Google and noticed something: the top organic result and the top paid ad were both for GitLab — the same company, buying its own branded keyword. He thought: what is the point of the ad layer? Who is this actually for?
Then his daughter started elementary school in the Bay Area and got a Chromebook, as Bay Area kids do, pre-installed with Google's full suite. He realized she would be tracked and profiled from age six to adulthood. A decade of childhood browsing behavior, owned by the same company selling access to it. You could pay for YouTube Premium to protect her from ads. There was no way to pay to protect her from being profiled in search. The option didn't exist.
So he built it.
The first five years ran on his own money — roughly $3M. Kagi went through a command-line prototype in 2019, a failed first attempt at paid search called Donna.gg, and an extended private beta before launching publicly in June 2022. Three months after launch, thousands of people were paying. The team hit 15 that year, entirely through word of mouth with zero marketing.
In 2023, the pricing shift changed everything. Kagi dropped from $14/month to $10/month for unlimited searches. That single move drove the subscriber count from ~5,000 to 20,000 in months. Simple pricing beat clever pricing.
The Marketing Playbook: Nothing
Kagi has never bought an ad. In the first three months of the public beta, Kagi had 2,600 paying customers and had sent exactly zero marketing emails. Prelovac wrote: "We believe that being bombarded with marketing emails the moment you sign up for a service is just not decent." That ethos held. Three years later, it still holds.
Growth came through Hacker News threads, Reddit posts, and Twitter arguments about Google's declining quality. When someone complained about AI Overviews pushing real results off the page, someone else recommended Kagi. The pattern repeated thousands of times.
One moment in 2023 captures how the flywheel works. Kagi's billing team quietly added a policy: if you don't use Kagi in a given month, you won't be billed. No press release. It appeared in the changelog. A few users posted it to Hacker News. Then it spread. That day became their biggest signup spike in six months. The "no use, no pay" rule cost almost nothing to implement — very few subscribers forget to use a daily product — and earned more trust than any ad campaign.
When Kagi raised $670K from 42 investors in 2023, most of them were already paying Kagi users. The round created people with a financial interest in Kagi growing who also happened to post on Hacker News. The investment structure became part of the acquisition funnel.
How Kagi Makes Money
| Plan | Price/month | What's included |
|---|---|---|
| Starter | $5 | 300 searches + standard AI models |
| Professional | $10 | Unlimited searches + Kagi Assistant (standard models) |
| Ultimate | $25 | Unlimited searches + all premium LLMs (Claude 4.5, GPT-5.5, Gemini 3 Pro) + Kagi Research deep research agent |
| Family (6 users) | $20 | Professional for up to 6 members |
| Teams | Per active user | Billed only for active users; no idle seat fees |
Annual plans are 10% off. That's not the interesting pricing detail. The interesting detail is the fair pricing commitment: Kagi will not bill you for a month you don't use the product. No VC would sign off on that. Kagi's user-investors did.
The Global Landscape: Private and Paid Search
Google owns roughly 90% of global search. Everything else competes for the remaining 10%, and within that slice, very different bets are being placed on what comes next.
The Big Players
| Company | Model | Scale | Revenue | What They're Betting On |
|---|---|---|---|---|
| Ad-funded | 8.5B searches/day | $175B+/year | AI Overviews keep users inside Google's ecosystem and reduce traffic to external sites | |
| Bing | Ad-funded + AI | ~4.3% market share | Part of $100B+ MS cloud revenue | Copilot integration makes Bing the default AI entry point for Windows users |
| DuckDuckGo | Privacy-friendly ads (Bing-backed) | ~100M searches/day | $100M+/year | Privacy-curious users who won't pay are still monetizable via contextual ads |
The Mid Players
| Company | Model | Scale | Revenue (est.) | Notes |
|---|---|---|---|---|
| Brave Search | Ad + Brave Rewards (independent index) | 1.6B queries/month | $100M+ ARR | One of only three independent indexes globally; powers Claude and Mistral web search |
| Perplexity | AI-answer-first, subscription | ~100M MAU | Undisclosed | $165M raised; dropped ad experiments in early 2026; $20-200/month plans |
| You.com | AI-agentic, multi-model | Undisclosed | Undisclosed | Enterprise focus; agentic workflow tooling |
| Startpage | Privacy proxy on Google results | ~3M MAU | Undisclosed | Shows Google results without passing user data to Google |
The Small Players
| Company | Model | Scale | Notes |
|---|---|---|---|
| Kagi | Subscription-only | 66,000+ subscribers | Profitable; zero VC; 50 people |
| Mojeek | Independent index, ad-free | ~9B page index | UK-based; one of three fully independent indexes globally |
| Ecosia | Ad-funded, plants trees | ~15M MAU | Contextual ads via Bing; plants trees with proceeds |
| Neeva | Subscription | Shut down | $77.5M raised; consumer product closed June 2023; sold to Snowflake for ~$185M |
The Neeva comparison matters here. Neeva raised $77.5M, hired the executive who had run Google's $115B ad business as CEO, and ran the same paid-search thesis Kagi is running. Ramaswamy's post-mortem was direct: "acquiring users has been really hard." Neeva had venture-scale costs and venture-scale expectations. Kagi had a founder's capital and a bootstrapper's patience. Same market. Different survival conditions.
How We Got Here
| Year | What Happened | Key Event |
|---|---|---|
| 2018 | Vladimir Prelovac founds Kagi | "Matrix moment" watching Google; daughter's Chromebook motivates the mission |
| 2019 | Donna.gg launches and fails | First paid search attempt abandoned; lessons carried forward |
| 2019 | Kagi Search prototype | Command-line prototype; early validation |
| 2020-21 | Private beta | Team grows from 3 to 10; search and Orion browser developed in parallel |
| Mar 2022 | Kagi Inc. incorporated; public beta launches | Thousands paying within 3 months; team grows to 15 |
| 2023 | $670K raised from 42 user-investors | Community angel round; crosses $1M ARR; team at 30 |
| 2023 | $10/month unlimited plan launches | Subscriber count goes from ~5K to 20K+ in months |
| 2024 | Profitable; PBC status; $2.5M total community raised | Kagi Assistant exits beta; team grows to 40 |
| 2025 | 50K+ subscribers; 1M+ daily searches | Kagi Research hits #1 SimpleQA benchmark at 95.5%; Orion exits beta; team at 45 |
| 2026 | 66K+ subscribers; Search API opens publicly | Team at 50; Kagi Translate app ships; aggressive AI product roadmap |
The Pattern You Can Steal
1. Remove the conflict of interest and let quality compound. Google's results are not bad because Google engineers are lazy. They're constrained by a business model that rewards clicks and penalizes quality that doesn't monetize. When you charge users directly, the incentive flips. Every product decision becomes: "does this make the product better for the person paying?" Kagi's growth is what happens when your business model and your product goal are the same thing.
2. Fair pricing as a viral moment. "No use, no pay." No marketing budget. One changelog entry. Their biggest single-day signup spike in six months. Trust is a compounding asset. It's almost impossible to buy with ad spend and almost impossible to lose once you've been genuinely fair with people. The lesson isn't "add a no-billing policy." It's: what would a fair neighbor do? Then do that.
3. Let your investors be your distribution. Kagi raised $670K from 42 users. Those users had a financial stake in telling their networks about Kagi. They posted about it on Hacker News. They recommended it on Reddit. They brought in colleagues. Community rounds in bootstrapped companies aren't just capital. They turn the people most likely to evangelize your product into people who also profit when it grows.
Quick Hits
Brave Search crossed 1.6 billion queries per month and now powers the web search layer inside Claude and Mistral's Le Chat. If AI companies keep outsourcing web retrieval to privacy-first indexes, Brave is quietly becoming the most important search infrastructure company most developers have never heard of. Brave Blog
Perplexity dropped its ad experiments in early 2026 and committed to a subscription-first model with tiers from $20 to $200/month. That's three major AI search players (Kagi, Perplexity, Anthropic) moving away from advertising as their primary revenue model, with OpenAI moving in the opposite direction. The market is now running a real-world experiment on which model wins which segment. eMarketer
Google's AI Overviews reduced click-through rates to external websites by an estimated 30-40% across several content categories in 2025. For publishers, that's existential. For Kagi users who pay specifically to reach original sources, it's a growth tailwind. Every point of friction Google adds to the open web makes Kagi's value proposition clearer.
What We're Watching
The AI bundling play. Kagi's $25/month Ultimate includes access to GPT-5.5, Claude 4.5, Gemini 3 Pro, Grok, and a deep research agent, all with zero training data usage. Perplexity's comparable offering runs $20-$200/month. If Kagi successfully repositions from "paid search" to "private AI access with great search included," the addressable market expands by an order of magnitude without changing the product.
The Bing API shutdown and what follows. Microsoft retired the Bing Search APIs in August 2025, cutting off a data source smaller search engines relied on. Kagi was affected. They've been public about it and are investing in their own Teclis index. If the DOJ's Google antitrust ruling eventually mandates index access on fair terms, Kagi would be positioned to benefit more than almost any other company in this space.
The path from 66K to 5M. Prelovac has stated publicly: reach 5 million customers (roughly $1B ARR), do it with a team under 75 people, do it in 5-9 years. That requires 75x subscriber growth while keeping headcount under 2x. The organic engine has worked so far. The question is whether word-of-mouth alone scales past a million, or whether Kagi eventually needs to market itself to a less tech-native audience — which would require a very different kind of machine from the one that got them here.
Ship it.
— The FounderSpec Team