The Search Stack War: Why Seltz’s $12.5M Bet on Machine-Ready Web Data Beats Google’s API Dependency

Posted on by

(SeaPRwire) –

By: Lucas Caldwell

[Paragraph 1]
The narrative that AI agents simply need better search is fundamentally flawed. It ignores the architectural reality of how machines consume information versus humans. Traditional search engines optimize for human attention spans and click-through rates. They serve snippets designed to entice a user to visit a website. This approach is obsolete for autonomous agents. Agents require machine-readable precision. They need structured data, tables, and specific textual contexts that reside deep within webpage bodies. The current reliance on generic APIs creates a bottleneck. It forces AI models to parse unstructured noise. Seltz understands this gap. Their funding round signals a shift toward owning the entire retrieval stack. This is not just another search wrapper. It is a foundational rebuild of how digital information is indexed for non-human readers. The old methods fail because they were built for eyeballs, not algorithms.

[Paragraph 2]
Seltz has secured $12.5 million in seed funding. Speedinvest and B Capital led the round. Founders Fund Italy, United Ventures, and Bain & Company’s Future Back Ventures participated. This capital injection supports a lean team of fifteen people. Only six are full-time employees. The company operates remotely across San Francisco, Pisa, and Leipzig. Founder Antonio Mallia brings deep expertise. He holds a PhD from NYU. He previously worked at Amazon’s AGI team and Pinecone. His vision challenges the status quo. He argues that transformer models require different inputs. The data needed for reasoning lies in tables and images. Snippets are insufficient for complex queries. Agents fire dozens of parallel requests. They need citations, not summaries. This distinction drives Seltz’s technical strategy. They aim to score individual passages. They extract specific elements for LLM consumption. This process is termed context engineering. It prioritizes precision over broad relevance.

[Paragraph 3]
The industry landscape reveals significant fragmentation. Many AI products rely on existing giants. OpenAI and Perplexity are building indexes. Anthropic and Mistral use Brave’s data. Google dominates through scale. Its billions of daily users provide unique visibility. However, this dependency creates vulnerability. Google recently sued SerpApi for scraping. This legal action highlights the tension between third-party tools and platform owners. Seltz offers an alternative path. It owns the crawler, index, retrieval models, and ranking system. This vertical integration allows for tighter control. It bypasses the limitations of public APIs. The system processes hundreds of millions of pages daily. Results return in under 200 milliseconds. This speed is critical for real-time agent workflows. Competitors like Parallel and Exa raise larger sums. Parallel secured $100 million. Exa raised $85 million. Tavily was acquired for up to $400 million. Yet Seltz bets on architecture over volume. Controlling the stack may prove decisive. It avoids the friction of middleware dependencies.

[Paragraph 4]
The race to build the best search infrastructure for AI is intensifying. Seltz’s approach focuses on deep structural integration. They target the specific needs of research agents. These agents demand high-fidelity data sources. The funding will support enterprise sales efforts. It will also expand the hiring of information retrieval experts. Many team members hold PhDs. They come from academic labs and major tech firms. This talent density suggests serious technical ambition. The market is crowded. Valuations are high. Competition is fierce. Success depends on execution speed. It relies on delivering superior query resolution. The era of human-centric search is ending. The age of agent-native retrieval is beginning. Companies that ignore this shift will lose relevance. Seltz is positioning itself at the center of this transition. Their strategy is clear. Build the stack. Own the data. Serve the machine.
Author bio: Lucas Caldwell, a tech opinion leader with millions of followers on X/Twitter