Onton, a San Francisco-based search and discovery company, has unveiled Ontology 1, a neurosymbolic model designed for complex, conversational, and multimodal product search. In a benchmark evaluation using 90 queries scored by three independent LLM judges, Ontology 1 achieved a mean precision@10 of 0.630, significantly outperforming Google Shopping (0.543) and Amazon (0.469)—a 2.7x improvement in accuracy—while indexing only about 1% of their catalogs.
Is It Deployable?
Yes, but not as downloadable weights. Ontology 1 is currently live for end users at Onton.com, and partner access is granted on a case-by-case basis for teams building on the agentic web. There is no public API, pricing tier, or open checkpoint. Adoption today means forming a partnership rather than running a simple pip install.
- Company Fit: Mid-market and enterprise retailers, marketplaces, and agentic-commerce platforms that struggle with long, requirements-heavy queries. Smaller catalogs may see less benefit, as the target failure mode scales with catalog size and listing noise.
- Industries: Currently focused on home decor and furniture, the only vertical Onton indexes. Onton claims the methodology generalizes beyond e-commerce, and that Ontology can search non-product data with minimal reconfiguration.
- Applications: Conversational and multimodal site search, moodboard-driven discovery, negation-heavy filtering, listing and review trust scoring, and grounding layers for shopping agents.
Why Keyword and Vector Retrieval Fail Here
Traditional e-commerce assumes intent maps onto categories and attributes—size, price, material, brand. There’s no filter for “pet-friendly” or furniture that fits your room. Onton argues this catalog interface has barely evolved in nearly three decades.
Ontology 1 addresses these gaps by integrating symbolic reasoning with neural retrieval. This allows the system to handle nuanced queries that combine visual, textual, and contextual cues. For instance, a query like “a comfortable sofa for a small apartment with pets” requires understanding material durability, space constraints, and lifestyle preferences—factors beyond typical attribute filters.
In 2026, as agentic commerce grows, users increasingly expect search to understand intent rather than match keywords. Ontology 1 positions itself as a foundational layer for this shift, enabling more intuitive and reliable product discovery across complex, multimodal inputs.
via MarkTechPost