As AI agents take on increasingly complex tasks, a growing question is how those agents find information on the open web. A startup called Seltz has raised a $12.5 million seed round to build search infrastructure designed specifically for artificial intelligence rather than for people.
The round was led by B Capital and Speedinvest, with participation from Future Present, Italian Founders Fund, Arc Investors, United Ventures, Vento Ventures, Mango Capital, 2100 Ventures, and Future Back Ventures, according to a report by Pulse 2.0 on June 25, 2026. Angel investors from Google, Ramp, Tako, and Hugging Face also joined. For more context on this story, see our ongoing artificial intelligence updates.
Seltz's central premise is that agents search the web very differently from humans, and that existing search engines, optimized for typed keywords and ranked result pages, are a poor fit for software that works on a company's behalf.
Why agents need a different search engine
Seltz argues that AI agents submit longer queries, run many searches in parallel, and require more complete context from the open web than a human typically needs. A person scans a list of blue links and picks one; an agent needs the full underlying document so it can reason over the content without making extra trips back to the web.
Rather than wrapping an existing search engine, Seltz is focused on owning more of the search stack end to end, building its own indexes and retrieval system. The company said its platform returns full documents rather than the previews that traditional search results offer. The goal is to help agents reduce extra retrieval steps and improve what has become a key cost concern for developers: token efficiency.
The startup also builds specialized indexes tuned to particular categories, so that each vertical can be optimized to the structure of the information being searched. It supports deployment in a customer's own infrastructure, a private cloud, or a required data region, and the company said it is SOC 2 Type II compliant, an important credential for enterprise customers in regulated industries.
Strong performance claims and an eight-month index
Seltz said it shipped a news index within eight months and has demonstrated what it calls strong performance on its own Dynamic News Search Benchmark. The company reported that its system answered at 89% accuracy and returned results in under 250 milliseconds.
The company also claims its own index surfaces documents that differ from those returned by Google-based systems, giving agents broader coverage when multiple search sources are used together. That differentiation matters in an emerging market where developers are increasingly combining several retrieval providers behind a single agent.
Built by search veterans
Seltz was founded by Antonio Mallia, who previously spent more than a decade working on web search, both as an academic researcher and at Amazon. The company currently has a 15-person team split across the United States and Europe. Most team members hold PhDs or have completed postdoctoral work in information retrieval, and several previously worked on Amazon's search teams.
That depth of retrieval expertise is part of the investor thesis. Building a web-scale search index from scratch is capital-intensive and technically demanding, and Seltz's backers are betting that a team steeped in information retrieval can build something purpose-built for agents that general-purpose search providers, who remain optimized for human behavior and advertising, will struggle to match.
Scaling to tens of billions of documents
Seltz plans to use the new funding to scale its indexes to tens of billions of documents and to expand the team needed to bring its search platform to more companies. On the engineering side, the company intends to grow from a limited set of verticals toward the broader open web while supporting on-premises and in-region deployment. On the commercial side, it plans to build out a team focused on reaching engineers and enterprises building agentic workflows.
The startup began commercial work late in 2025 and currently has a foundational research lab under contract, along with pilots running with teams building agentic applications.
A bet on agentic infrastructure
Seltz is entering a category that is attracting serious investor attention. As companies move from chatbots to autonomous agents that take actions on their behalf, the infrastructure those agents rely on, tools, memory, and retrieval, has become its own market. Web search is among the most fundamental of those building blocks, because an agent that cannot find current, accurate information is unlikely to be useful.
The competition is already forming. Larger search and cloud providers are racing to add agentic capabilities, while a wave of startups is targeting specific layers of the agent stack. Seltz's wager is that retrieval built from the ground up for machines, not humans, will become a category of its own.
With $12.5 million now in hand and a team of retrieval specialists, the company has the runway to test that thesis. Whether agents ultimately need their own search engine, or whether the incumbents adapt fast enough to serve them, is one of the open questions defining the next phase of the AI infrastructure boom.
Source: "Seltz Raises $12.5 Million Seed Round To Build Web Search For AI Agents," Pulse 2.0, June 25, 2026; Fortune.---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
