Keenable Bets Big on AI-First Search With $26 Million Seed Round

Keenable has emerged from stealth with $26 million in seed funding to build search infrastructure designed for AI agents rather than human web surfers. The startup, backed by Accel, says its index already spans more than 100 billion documents and is being used in production by AI labs and inference providers as the web shifts toward machine-driven search and task completion.

Founded by former Yandex search chief Andrey Styskin and AI scientist Matthias Petri, Keenable is positioning itself as a core layer for the next generation of chatbots and autonomous assistants. Its pitch matters because the companies that power search for AI systems are becoming just as important as the models themselves.

Why Keenable thinks search needs to change for AI

Keenable’s central argument is straightforward: the internet was organized for people, but increasingly it is being read by machines. Human users usually scan a few results, skim a page and move on. AI systems, by contrast, can ingest far larger amounts of material and combine information from multiple sources before generating an answer.

That difference, according to the company, creates a new kind of search problem. Instead of optimizing for quick clicks and page views, infrastructure now has to help bots retrieve relevant source material fast enough, cheaply enough and reliably enough to support real-time AI applications.

How is AI search different from traditional search?

AI search is different because it must support grounding, retrieval and synthesis across large document sets, not just ranking links for a human. In practice, that means the system has to find precise evidence, narrow huge search spaces quickly and deliver data that can be used immediately by a model during training or inference.

Styskin has argued that AI systems perform better when their answers are backed by original documents, because those sources reduce hallucinations and improve the quality of generated responses. He described this as a feedback loop unlike the one Google built around human search behavior.

Styskin said the company sees a different search economy emerging for AI systems, one in which grounding responses in documents creates a flywheel unlike traditional web search.

Who is behind Keenable?

The company is led by two founders with deep technical search backgrounds. Styskin previously ran Yandex’s search, AI and cloud division, giving him extensive experience building systems for large-scale query processing. Petri is a German AI scientist, and the pair are combining search infrastructure expertise with machine-learning know-how.

That background is one of Keenable’s biggest selling points. Search is not a new category, but the founders believe the technical demands of AI-era retrieval are different enough to create room for a fresh platform built from the ground up.

What experience are the founders bringing to the table?

They are bringing years of work on search infrastructure at both Yandex and Amazon. Styskin said his time building systems for AI applications, including work linked to Alexa at Amazon, helped him see how retrieval should be adapted for agentic software rather than consumer browsing.

He also pointed to traffic trends in the broader web ecosystem, including Cloudflare data showing that AI crawlers are taking up a larger share of search activity. For Keenable, that is evidence that bots are already reshaping demand even before the market fully adjusts.

The funding round and what it signals

Keenable’s $26 million seed round was led by Accel, with participation from Conviction Partners and several angel investors. The size of the round is notable for a company that has only recently come out of stealth, reflecting investor appetite for infrastructure that supports the rise of AI agents.

The backing also suggests that venture firms view AI retrieval as a foundational layer in the stack, not a niche feature. As models become more capable of acting on behalf of users, the companies that help them search, verify and retrieve information could become indispensable.

How does Keenable say it stands apart from enterprise search?

Keenable says it is not simply repackaging enterprise search for corporate knowledge bases. The startup argues that web-scale retrieval is a much harder problem because the internet is enormous, constantly changing and expensive to scan in full.

According to Styskin, generic indexing systems become inefficient when they are stretched across the entire web without task-specific tuning. His view is that the winning product must quickly reduce the search space based on the query, instead of brute-forcing results at great cost.

That focus on speed and selectivity matters because AI agents may need to query the web many times in a single task. If each query is too expensive or too slow, the entire economics of agentic software break down.

Why is cost such a major issue?

Cost is a major issue because web-scale indexing and retrieval consume significant compute and storage resources. Styskin has acknowledged that building and operating a large search index is extremely expensive, but he argues that the company is structuring its system carefully to manage those costs over time.

For AI companies, the economics of retrieval matter as much as accuracy. A search layer that is technically excellent but too costly to serve may never be practical for widespread deployment.

Styskin said the economics of serving the entire internet make cost control difficult, but he believes better index design can narrow the problem enough to make the system viable.

Why Google and Microsoft matter to Keenable

Keenable’s opportunity is tied in part to a change in strategy by the largest search players. Accel partner Zhenya Loginov, who led the investment, said companies building AI products have fewer choices for web-scale search infrastructure because Google and Microsoft have moved to restrict or shut down some search APIs that previously supported developers.

Those changes reflect a broader conflict between platform control and partner access. The major incumbents are increasingly bundling search into their own products and choosing partners more selectively, which leaves room for independent infrastructure companies to step in.

For startups building chatbots, assistants and other AI-driven products, that can translate into a search dependency problem. If the big platforms become less open, alternative providers may gain leverage even if they are much smaller than the incumbents.

What products is Keenable building?

Keenable says its current offering includes a web search index and an API used by AI labs and inference providers in production. The company has not identified those customers publicly, but it says the usage covers both training and runtime workflows.

The startup is also developing proprietary retrieval technologies. One upcoming product, called WebQueryLanguage, is intended to let AI systems answer more complex questions by combining material from multiple web sources when no single page contains the full answer.

That kind of multi-source synthesis is increasingly important in AI search. Many useful answers are not located in a single document, and the system has to infer relationships across pages, datasets and references while still preserving traceability.

What is WebQueryLanguage supposed to do?

WebQueryLanguage is designed to help AI systems merge information from several web sources into one answer. The idea is to support queries where evidence is scattered and no single result is complete on its own.

If it works as promised, the product could help bots produce more complete responses for research, voice assistants and agent workflows that need to reason over fragmented information.

How are AI labs already using web search infrastructure?

AI labs are already using search infrastructure in multiple parts of the model lifecycle. During training, search can help systems find relevant corpora, label data and improve retrieval-augmented learning. During runtime, it allows chatbots and agents to look up fresh information and ground answers in current sources.

Keenable says its API is being used in both settings, which suggests the company is trying to become part of the underlying plumbing for AI development. That is a valuable position if the market continues to shift from static model outputs toward live, evidence-based responses.

The growing importance of runtime retrieval is one reason investors are paying attention. As models become more capable, users will expect them to cite, verify and act on information in real time rather than rely solely on pretraining.

What competitive landscape does Keenable face?

Keenable is entering a field that already includes several ambitious players. Brave and Exa are among the companies building alternative search tools with AI use cases in mind, and Google itself is redesigning parts of search for the generative era.

That means Keenable is not alone in believing that search is being rewritten. The question is whether it can carve out a durable role as a specialist infrastructure provider rather than merely a feature supplier.

The startup’s advantage may lie in its focus on the lowest layer of the stack. If it can prove that its index is faster, more selective and more economical for agent queries, it could become a behind-the-scenes standard for AI developers.

What does the rise of AI crawlers mean for the web?

The rise of AI crawlers suggests the web is changing from a human-first information network into a dual-use environment for both people and machines. That shift has implications for publishers, infrastructure providers and search companies alike.

As more AI systems crawl, summarize and answer questions directly, the old logic of sending users to ten blue links may weaken. If users get what they need from a generated response, search becomes less about directing traffic and more about supplying verified context to an agent.

This transition could reshape business models across media and software. Sites that once depended on referral traffic may need to rethink how they package information, while search providers may have to support machine-readable access patterns instead of human browsing habits.

Why investors see an opening now

Investors appear to believe the timing is right because the market is moving faster than incumbents can comfortably adapt. Google and Microsoft may still dominate search, but their incentives are tied to protecting their own product ecosystems, not necessarily to supplying every AI company with open access.

That creates a gap for a specialized vendor that can sell flexibility, neutrality and lower-cost retrieval. If AI agents keep spreading into consumer and enterprise software, the demand for independent search infrastructure could rise along with them.

Keenable’s funding is also a signal that technical credibility matters in this market. Investors are not just betting on a business model; they are betting on the founders’ ability to solve an infrastructure problem that has resisted easy answers for decades.

What happens next for Keenable?

The company says it plans to use the new capital to expand its team from 15 engineers to roughly double that size by the end of the year, with an emphasis on go-to-market hiring. That suggests Keenable is moving beyond pure product development and into commercial scale-up mode.

For now, the startup still has to prove that its technology can hold up under real customer demand and still remain cost-effective. The challenge is not simply to build a massive index, but to operate it efficiently enough that AI companies choose it over both incumbents and newer rivals.

Whether Keenable can become, as it hopes, “the next Google for AI agents” will depend on execution, adoption and the pace of the broader shift toward agentic software. But the thesis behind the company is already clear: if machines are becoming the primary consumers of the web, the internet’s search layer may need to be rebuilt for them.

Key milestones in Keenable’s launch

The company’s path from stealth to public visibility has been rapid, and the timeline helps explain why the startup is attracting attention from both investors and AI builders.

Keenable’s debut highlights a bigger industry shift: AI is not just changing what people search for, but how search itself has to work. The startups that can make web retrieval faster, cheaper and more accurate for agents may become essential parts of the next internet layer.