Seltz Raises $12.5M Seed to Rebuild Web Search for AI Agents

Search infrastructure for AI agents is becoming its own competitive market, and Fortune reports that Seltz just raised $12.5 million in seed funding to take a run at it. The round was led by Speedinvest and B Capital, with additional backing from Italian Founders Fund, United Ventures, and Future Back Ventures, the venture arm of Bain & Company. The premise is that traditional search was designed for humans skimming ranked links, while AI agents fire off long, precise queries at scale and need machine-readable information they can cite directly.

What Seltz is betting on is full-stack ownership. Rather than wrapping Google, Bing, or Brave APIs, the company built its own web crawler, search index, retrieval models, and ranking pipeline. According to the reporting, the system crawls hundreds of millions of pages a day and returns results in under 200 milliseconds, scoring individual passages and extracting specific tables, text, or images rather than returning full pages.

CEO Antonio Mallia holds a PhD in computer science from New York University with a focus on information retrieval, and previously worked as an applied scientist on Amazon's artificial general intelligence team and as a research scientist at the vector-database company Pinecone. He told Fortune the moment reminds him of when Google's PageRank upended search in the early 2000s, saying "The revolution is back again," driven this time by transformer models and the AI workflows that increasingly do the searching themselves.

The competitive pressure is real and the funding asymmetry is significant. Parallel, founded by former Twitter CEO Parag Agrawal, has raised $100 million at a $2 billion valuation. Exa has pulled in $85 million. Tavily was acquired by the AI-cloud company Nebius for up to $400 million. Against that backdrop, Seltz's $12.5 million is a starting position. What the reporting does not give you is any revenue figure or named customer, and the sub-200ms performance claim is one that will need to prove itself under real enterprise load and genuine retrieval quality tests.

If the full-stack approach holds up, the clearest beneficiaries are developers building research agents, autonomous pipelines, and AI-powered workflows who currently depend on search infrastructure that was never designed for machine consumption. A retrieval layer purpose-built for agents could become a foundational piece of the broader AI infrastructure stack.