Keenable’s AI Index: A New Bottleneck for Foundational Knowledge?

The Illusion of Decentralized AI Access

As Keenable secures a hefty $26 million in seed funding to build a web search index specifically for AI agents, the industry consensus is that this fills a critical gap. Current search engines, optimized for human interaction, fail to serve the voracious, nuanced data demands of autonomous AI. Keenable, founded by former Yandex and Amazon search veteran Andrey Styskin and German AI scientist Matthias Petri, aims to provide AI systems with a more efficient, cost-effective way to ground their responses in reliable, web-scale data, boasting an index of over 100 billion documents.

This initiative, backed by Accel, is presented as an essential innovation, born from the reality that tech giants like Google and Microsoft are increasingly restricting access to their search APIs. Accel partner Zhenya Loginov notes the scarcity of web-scale infrastructure for AI. Styskin himself points to Cloudflare data showing a surging share of search volume originating from AI crawlers, framing Keenable as a response to this undeniable shift. But beneath the surface of technical necessity lies a deeper, more unsettling implication: the very infrastructure intended to liberate AI’s access to information risks creating a new, proprietary gatekeeper.

Re-Centralizing the Web for Agents

The narrative frames Keenable as a disruptor, a nimble player capitalizing on Google’s ‘innovator’s dilemma.’ Styskin suggests Google is “beatable” on agentic queries, acknowledging it is “extremely hard” to dislodge the incumbent for traditional search. Yet, Keenable’s ambition — to become “the next Google for AI agents” — explicitly seeks to replicate the very centralized power structure it purports to bypass. This is where the industry’s collective optimism deserves a dose of skepticism. Instead of fostering a truly open and federated knowledge base for AI, we are seeing the foundational layer for future AI intelligence being funneled through new, privately owned choke points.

What incentive do venture capitalists and startups have in this framing? Simply put, the promise of a defensible, proprietary data moat. Building such an index, as Styskin candidly admits, is “painfully expensive.” This cost barrier naturally limits competition, creating a winner-take-all dynamic familiar from the early internet. Accel and Keenable benefit immensely from positioning this as an urgent, technical solution, obscuring the long-term structural implications for information access and AI development. The capital deployed isn’t just for technology; it’s for securing future leverage.

The Long Shadow of Search Monopoly

The irony is profound. A significant portion of the AI community champions decentralization and open access, yet the practical demands of training and operating large language models (LLMs) and agents push them towards specialized, centralized data providers. When Google and Microsoft withdraw their public search APIs, they effectively force the ecosystem to either build their own infrastructure – an insurmountable task for most – or turn to companies like Keenable. This dynamic creates a vacuum that new players eagerly fill, but with a business model inherently geared towards exclusive control over the aggregated data.

Keenable’s upcoming Web Query Language, designed to combine information from various web sources, aims to help AI systems answer complex questions without a single definitive source. This is a powerful capability, but its power is inherently constrained by the underlying index. If only a few entities control the most comprehensive and high-quality indices, they effectively dictate the ‘worldview’ available to future AI agents, subtle biases and all. The shift from “ten blue links” to AI-generated answers might seem like progress, but if those answers are uniformly shaped by a handful of commercial filters, the internet’s original promise of diverse information could be further diminished.

The Cost of AI’s Knowledge Foundation

The economic realities of building web-scale infrastructure are undeniable. Styskin, drawing from 20 years at Yandex and Amazon, emphasizes the need for fine-tuned index structures to manage the “enormous” cost of scanning the entire internet. This technical challenge, however, directly feeds into the centralization trend. Only well-funded ventures can even attempt such a feat, leaving smaller, innovative AI labs increasingly dependent on these new data intermediaries.

As Keenable grows its 15-person engineering team, aiming to double headcount, its focus is clear: refine proprietary retrieval capabilities and scale. But the question remains whether this scaling ultimately serves a broader, open AI ecosystem or entrenches a new form of data oligopoly. The history of the internet is replete with open protocols giving way to closed platforms. The critical lesson for AI is to not let the pursuit of technical efficiency inadvertently pave the way for a knowledge bottleneck that stifles innovation and limits the collective intelligence of the agentic web before it even truly begins.