Ontology 1 by Onton Outscores Google and Amazon on Search Precision

What happened

San Francisco-based search and discovery company Onton has launched Ontology 1, a neurosymbolic model designed for complex, conversational and multimodal product search in e-commerce environments. The release marks Onton's first publicly benchmarked model and positions the company directly against the dominant search infrastructures operated by Google and Amazon.

On a 90-query benchmark evaluated by three independent large-language-model judges, Ontology 1 recorded a mean precision@10 score of 0.630, compared with 0.543 for Google Shopping and 0.469 for Amazon. Notably, Onton achieved these results while indexing approximately 1% of the product catalogue volume used by those platforms, suggesting the model's advantage lies in reasoning quality rather than raw data scale.

Why it matters

Search is one of the highest-stakes moments in the e-commerce customer journey. When a shopper types or speaks a complex, nuanced query — "a lightweight waterproof jacket for hiking in humid climates under £150" — the gap between a relevant result and an irrelevant one is the gap between a conversion and an abandoned session. Precision at the top of a results page is therefore a direct proxy for customer satisfaction, trust and repeat purchase intent. Ontology 1's neurosymbolic architecture, which combines neural pattern recognition with symbolic reasoning, is specifically built to handle the kind of layered, conversational intent that keyword-matching and standard vector search routinely mishandle.

From a behavioral-economics perspective, this matters because choice overload and poor result relevance are well-documented drivers of decision fatigue and cart abandonment. A search layer that surfaces fewer but more precisely matched products reduces cognitive load, shortens the path to purchase and reinforces the perception that a retailer "understands" the customer — a powerful trust signal that compounds over time.

By the numbers

• 0.630 — Ontology 1's mean precision@10 score on the 90-query benchmark
• 0.543 — Google Shopping's score on the same benchmark
• 0.469 — Amazon's score on the same benchmark
• ~1% — the approximate share of catalogue volume Onton indexed relative to the competing platforms
• 3 independent LLM judges used to score benchmark results

The Renascence take

The headline comparison to Google and Amazon will dominate coverage, but the more instructive detail is the catalogue-size asymmetry. Onton outperformed on precision while working with a fraction of the data — which points to a structural shift in how search quality should be measured and procured by retailers.

Most e-commerce operators still treat search as an infrastructure commodity, benchmarking vendors on catalogue coverage and query speed. Ontology 1 reframes the conversation around intent fidelity — how faithfully a model interprets what a customer actually means, not just what they typed. The behavioral principle underneath is straightforward: customers do not experience your index size, they experience the relevance of the first ten results. Operators evaluating search vendors in 2025 and beyond should weight precision-at-small-k metrics far more heavily than they currently do, and pilot neurosymbolic or hybrid-reasoning approaches on their highest-abandonment query segments first.