Seltz

Market position

Seltz competes in AI-native web search and retrieval infrastructure against better-funded entrants including Parallel, Exa, and Tavily, as well as incumbent search APIs from Google, Bing, and Brave that many AI search products resell. Its stated differentiation is operating an independent index rather than reselling incumbent results. Press coverage notes Google is still regarded as having the best index due to its scale.

The company positions ownership of the entire pipeline — crawler, index, retrieval models, and ranking — as its central differentiator, arguing that providers routing queries to Google or Bing return documents that substantially overlap with what an agent can already retrieve, so a second call adds cost without new information. Claimed benefits of the owned stack are result diversity, dynamic configurability, and latency roughly 7x lower than competing providers, alongside full-page rather than snippet-level context and relevance ranking tuned to agent tasks rather than human click behavior.

Technology

Seltz builds its full retrieval stack in-house: web crawler, index, retrieval models, and ranking. The company states the system crawls hundreds of millions of pages per day and responds in under 200-250 milliseconds, which it benchmarks as 7x-30x faster than other search API providers whose average response time it measured at approximately 1.2 seconds. Instead of returning full pages or summaries by default, it scores passages and extracts the relevant table, text, or image, a process the founder calls context engineering. The company argues existing open-source and commercial search tooling is designed for enterprise or small-scale search over thousands or millions of documents and cannot simultaneously be fast, cost-effective, and capable of web-scale indexing.

Go-to-market

The company combines a developer-led motion — free-to-start API keys, Python and JavaScript SDKs, documentation, a playground, and integrations with OpenAI, Anthropic, Mistral, Agno, n8n, and MCP-compatible IDEs and assistants — with a nascent direct enterprise sales effort funded by the seed round. It also publishes technical content and a public benchmark (DNSB) comparing retrieval quality and latency across AI search providers, and features endorsements from executives at Tako, Synthesia, and Cohere on its site.

Teams building AI systems that require live web knowledge — large language model applications, RAG pipelines, and autonomous agents — with go-to-market focused on frontier AI labs, AI-native startups, and enterprise companies.

Geography

Incorporated in the United States and described in press coverage as San Francisco-based, Seltz operates as a fully remote team split between the San Francisco Bay Area and hubs near European universities in Pisa, Italy, and Leipzig, Germany.

History

Seltz was incorporated in the United States and founded in October 2025, with the company describing itself as founded in 2025. It publicly released its web search API in April 2026, beginning with US news coverage, accompanied by a founder blog post explaining the decision to build the crawler, models, and retrieval stack from scratch. In June 2026 the company announced a $12.5 million seed round. It also launched the Dynamic News Search Benchmark (DNSB), a public benchmark for comparing web retrieval quality and latency across AI search providers.

Risks & controversies

Seltz enters a market with better-capitalized competitors — Parallel raised $100 million at a $2 billion valuation, Exa raised $85 million, and Tavily was acquired by Nebius for up to $400 million — while operating with roughly six full-time employees. Building and maintaining an independent web-scale index is capital- and engineering-intensive, and incumbent index scale remains an advantage for Google. The broader sector also faces legal friction over web data access, as illustrated by Google's December suit against SerpApi over alleged circumvention of anti-bot protections.