AI agents are changing the way machines interact with the web — and that is creating a demand that the current internet infrastructure simply cannot keep up with.
The internet was built with people in mind. Search engines like Google were optimized to deliver fast, relevant results for humans who have neither the time nor the patience to comb through entire pages. But chatbots and AI agents work in a completely different way: they can read, process, and cross-reference much larger volumes of information all at once. The problem is that the search infrastructure we have today was never designed for them, and a growing number of people believe the web needs an upgrade to serve these machines — not just human users.
That is exactly the gap Keenable wants to fill. The startup, backed by Accel, recently emerged from stealth mode with $26 million in seed funding and a pretty ambitious goal: build the equivalent of a Google, but designed from scratch to serve AI agents. 🤖 The round was led by Accel, with participation from Conviction Partners and several angel investors. With an index of more than 100 billion documents already indexed and an API actively in use by AI labs and inference providers, the company wants to become the default search layer for the next generation of intelligent systems. Sounds too ambitious? Maybe. But the names behind the project and the market timing suggest this bet could make a lot of sense. 👇
Who is behind Keenable
Keenable was founded by Andrey Styskin, who previously led the search, AI, and cloud division at Russian tech giant Yandex, and by German AI scientist Matthias Petri. Together, they bring more than two decades of experience building large-scale search systems at companies like Yandex and Amazon.
It was actually during their time at Amazon, building web search infrastructure for AI applications like Alexa, that the pair began to see the opportunity. Styskin noticed Cloudflare data showing that AI crawlers were responsible for a growing share of web search volume. That was the moment it clicked: there was room to build search infrastructure designed from the ground up with artificial intelligence in mind.
According to Styskin, AI chatbots tend to perform much better when they can anchor their responses in source documents. In his words, this creates an entirely new virtuous cycle, different from the one Google learned by observing human behavior over the years. In other words, the logic that powers search for machines is fundamentally different from what powers search for people.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
Why current search infrastructure does not work for AI agents
When you search for something on Google, the engine delivers a list of links in a fraction of a second. That process has been refined over entire decades to cater to a very specific behavior: a human being who wants a quick answer, clicks on one or two results, and moves on. The design behind it is brilliant for what it sets out to do, but it starts from a fundamental assumption — that there is a human user on the other side of the screen, with limited attention, scarce time, and a straightforward question to resolve.
Modern AI agents and chatbots break that assumption completely. They are not looking for a link to click. They need structured, continuous, and high-volume access to web data to fuel complex reasoning, provide in-depth answers, cross-reference sources, and build elaborate responses in real time. An AI agent might need to consult hundreds of different documents in a single task, something no current search engine can deliver efficiently and at scale for this type of use. The latency, the format of the data returned, and the ranking logic — everything was designed for humans, not machines.
This mismatch creates a real bottleneck in the development of more capable AI systems. When an agent needs up-to-date information from the web and the available infrastructure was not built to handle that kind of request at scale, the result is slowness, inaccuracy, or sky-high operational costs to work around the problem. It is a concrete technical obstacle — and that is exactly where Keenable thesis comes in.
What Keenable is actually building
The core proposal behind Keenable is easy to understand but complex to execute: build a web indexing infrastructure designed specifically to serve AI agents and chatbots. It is not an adaptation of something that already exists. It is an architecture built from scratch with AI system needs as the starting point, not as an afterthought. That changes quite a bit about how data is collected, organized, updated, and delivered.
With more than 100 billion documents already in its index, the company already has the scale to position itself as a serious search layer. The API developed by the startup is actively used by AI labs and inference providers, both during training and at runtime. The company does not reveal who its clients are, but it recently closed a partnership with Gradium, a voice AI company, to support live information retrieval. That kind of traction at seed stage is relatively rare and makes a real difference in the project credibility in the market.
Styskin explains that Keenable product is different from enterprise search solutions, which tend to break down and get extremely expensive when applied at web scale. According to him, if you do not fine-tune index structures for a specific task, the cost of serving and scanning the entire internet becomes massive because of the sheer volume. That is why you need to innovate on how to quickly narrow down the search space based on the query — and that is exactly what Keenable brings to the table.
From a technical standpoint, the difference between indexing for humans and indexing for AI agents goes far beyond volume. It involves how data is structured on delivery, how frequently the index is updated, the ability to process queries with richer semantics, and the capacity to return relevant context — not just links. The startup is also developing proprietary retrieval capabilities, including an upcoming product called Web Query Language. The idea is to help AI systems answer questions by combining information from multiple web sources, even when none of them contains the complete answer on its own. 🔍
The right timing and the giants dilemma
Timing matters a lot in this story. According to Zhenya Loginov, the Accel partner who led the investment, AI players have very few options when it comes to search infrastructure at web scale. That is because giants like Google and Microsoft have been taking steps to shut down their existing search APIs, precisely to avoid cannibalizing their own products. Instead, these companies are opting for a more closed approach and being quite selective with their partners.
For Styskin, those decisions confirmed the opportunity he had already been eyeing. He believes that, because of the classic innovator dilemma, Google is potentially beatable on agent-oriented queries. A smaller company like Keenable can innovate faster and offer a more cost-efficient solution for AI companies. He acknowledges that it is extremely difficult to convince people to abandon Google for traditional searches, but the landscape of agentic queries opens up a real window of opportunity.
Still, the costs of building a massive search index are very real. When asked about it, Styskin was blunt: do not ask, it is painfully expensive. Even so, the startup says it is doing everything it can to keep costs under control and grow at the right pace. With a team of 15 engineers spread across the United States and Europe, the company plans to use the new capital to double the team by the end of the year and build out its go-to-market strategy.
What this means for people building with AI
For developers and teams building products with chatbots and AI agents, the existence of infrastructure like what Keenable is proposing could significantly change the equation. Today, anyone who wants to give their agent access to up-to-date web information has to solve that problem in makeshift ways — using search APIs that were not built for this, building their own scrapers, or paying a premium for partial solutions that deliver less than what is needed.
An indexing API built specifically to serve AI agents simplifies that process in a meaningful way. Instead of spending time and resources developing and maintaining the information access layer, teams can focus on what truly differentiates their product: the agent logic, the user experience, the integration with other systems. It is the kind of infrastructure that, when it works well, fades into the background — you stop thinking about it because it just works. ⚙️
On top of that, the quality of the information feeding an AI agent has a direct impact on the quality of the answers it delivers. A larger, more up-to-date index with delivery optimized for how AI systems consume data means more accurate agents, more useful responses, and fewer hallucinations — one of the most persistent problems with large language models in production. In that sense, what Keenable is proposing is not just about technical infrastructure. It is about raising the bar for AI systems as a whole.
The race for a new web built for agents
Keenable is not alone in this journey. Other players have already entered the space, like Brave and Exa, each with their own vision of how to serve data to AI systems. And Google itself is redesigning its search experience for the AI era. This broader movement shows that Keenable conviction is shared even inside the Googleplex: whether for humans or for agents, the era of the famous ten blue links may be coming to an end.
Keenable is still in its early stages, and building the equivalent of a Google for AI agents is a massive challenge — technical, operational, and commercial. But the gap it identified is real, the market is growing fast, and specialized search infrastructure for artificial intelligence could be one of the most important pillars of the next phase of the web. Definitely one to keep an eye on. 👀
