Onton Releases Ontology 1: A Neurosymbolic Search Model That is 2.7x More Accurate than the World’s Best E-…

What it does

Onton, a San Francisco-based search company, has launched Ontology 1, a neurosymbolic search model designed for complex, conversational, and multimodal product searches. This model combines neural network capabilities with symbolic reasoning to handle nuanced queries better than traditional e-commerce search engines. Ontology 1 was benchmarked on 90 product search queries and evaluated by three independent large language models, achieving a mean precision@10 score of 0.630. In comparison, Google Shopping scored 0.543, and Amazon scored 0.469, marking a significant accuracy improvement.

Why it matters

Higher precision in search results matters because it directly reduces user friction in e-commerce, improving conversion rates and customer satisfaction. Ontology 1’s 2.7 times better accuracy compared to top platforms pushes the boundaries on how well machines understand complex, conversational product queries—typically a notorious weak spot. By indexing roughly 1% of the data volume Google and Amazon use, Onton demonstrates that smarter search design can outpace brute-force data scale. This challenges incumbent e-commerce giants who rely heavily on massive data and traditional search heuristics, creating pressure to innovate or risk losing shopper attention to more intelligent alternatives.

Who it is for

Ontology 1 targets product search teams and operators in e-commerce who struggle with complex, multi-turn customer queries that combine text and images. Builders and founders aiming to improve product discovery through conversational interfaces or visual search can consider integrated neurosymbolic approaches like Ontology 1. Retailers looking to boost shopper experience and reduce reliance on massive backend indexing may find this model attractive given its efficiency and higher accuracy.

The catch

Despite its promising accuracy, Ontology 1 currently indexes only about 1% of the data scale used by Google and Amazon. This raises questions about performance at massive scale and the ability to handle very large catalogs in real-time applications. Also, because the benchmark uses LLM judges rather than human evaluators, the practical commercial impact will depend on how well this translates into actual user behavior and revenue lift for clients.

What to watch next

Look for Onton’s next move to scale Ontology 1’s indexing capability or integrate it into live e-commerce platforms. Watch for third-party validations comparing its search effectiveness with human shopper satisfaction metrics. Competitors like Google and Amazon may respond by updating their models or incorporating similar neurosymbolic techniques, squeezing the advantage Onton currently holds. Finally, see if other verticals beyond e-commerce adopt this hybrid neural-symbolic search to tackle complex multimodal queries.

AI Quick Briefs Editorial Desk