Onton has raised $7.5 million to bring its AI-powered shopping platform beyond furniture, taking a bet that a neuro-symbolic approach to product discovery can outperform chat-only assistants and traditional search.
The startup, which rebranded from Deft earlier this year, has seen its monthly active users grow to more than 2 million from about 50,000 as it gets ready to add apparel next and consumer electronics after that.
Footwork led the round, with Liquid 2, Parable Ventures and 43 contributing to it, bringing Onton’s total funding to approximately $10 million. It comes as AI-first shopping startups and tech giants race to make browsing more conversational and visual, with OpenAI, Google and Amazon pushing assistants and open-source product research tools, while players like Perplexity, Daydream and Cherry work on discovery engines from the bottom up.
Funding and expansion plans for Onton’s AI platform
Onton’s new funding will allow it to expand categories, further grow its team and continue R&D. The company has expanded from three full-time employees in 2023 to 10 today, and it intends to have a staff of 15, including engineers and researchers. Leadership presented the rebrand as a pragmatic play to cut down on name confusion and land a stronger domain, an experience many consumer-facing platforms struggle with (a pain point known as trust and recall).
Apparel is the immediate goal, although an underwear catalog is already underway, followed by consumer electronics. But that path puts Onton in even noisier turf, where differentiation comes down to attribute-level knowledge, image-driven site browsing and the ability to reconcile messy/inconsistent product data across retailers.
Onton’s Neuro-Symbolic AI Behind the Scenes
Onton’s central pitch is that they are powerful in predicting likely answers, but subject to hallucinations and loose semantics which detract from their precision in commerce. This is tackled by Onton using a neuro-symbolic architecture, which integrates neural networks with explicit rules and ontologies to impose logical consistency. The hybrid approach has caught on in research settings, including at institutions like IBM Research and Stanford University, for tasks that require both perception and reasoning.
In practice, this means Onton can infer attributes missing from product descriptions and understand synonyms across retailers, or it can make purchases based on more verifiable signals.
Search “pet-friendly sofa,” for example, and you may see surfaces of polyester or tightly woven performance fabrics that repel stains and are scratch-proof, the result perhaps of a thought process that extends beyond simply matching keywords. The company says its system improves inference of attributes and understanding of search queries with each and every search and interaction over time.
Onton makes it possible with multimodal inputs. Shoppers can also upload a room photo, create design concepts on an infinite canvas, or merge inspiration images with products to iterate quickly. Instead of shoehorning everything into a chat window, the interface encourages navigation, editing and reordering — actions that accurately reflect how people plan when they are buying. The company says that its conversion is anywhere from “3–5x higher than typical e-commerce baselines,” which would be a very impressive figure if it’s true, given that the average conversion for many players in the space tends to be somewhere around single digits according to Adobe’s retail benchmarks and industry trackers.
Competitive landscape as Onton expands beyond furniture
Expanding into clothing will also put Onton in competition with AI-native shopping startups, such as Daydream, Aesthetic and Style.ai, as well as the traditional search and recommendation systems employed by large retailers. Meanwhile, discovery-first tools from Perplexity and Cherry, and assistant-driven experiences from Amazon and Google continue to squeeze the top end of the funnel (where shoppers do research, comparison and intent refinement).
Onton’s differentiators sit across three legs: structured reasoning to minimize hallucinations in recommendations, dense attribute mapping that unifies conflicting catalog data, and a canvas-style workflow that can supply longer, more collaborative purchase journeys. If those elements live through apparel and electronics intact as they did in furniture, then Onton may just be able to create a defensible niche for itself in AI-powered commerce.
Why this is important for retail and e-commerce performance
The drag on e-commerce performance remains product findability. The Baymard Institute measures cart abandonment at about 70 percent, because it says bad product search and weak decision support are some of the main reasons for dropping out. Analysts at McKinsey have made a similar argument around the potential for generative AI to unlock huge value in retail by perfecting discovery, personalization, and service; they’ve also warned that reliability and governance are what will make those gains or break them.
For AI shopping to work at scale, systems need to be transparent, grounded and tolerant of ambiguity — especially in categories where fit, finish and compatibility are important. Neuro-symbolic approaches are well suited to encode rules such as material durability, sizing equivalences and spec-matching rules into a model and then combine with learned embeddings from text and image bases to cover the long tail of queries.
What to watch next as Onton moves into new categories
Now the big questions are around execution: how fast Onton can get high-quality catalogs beyond furniture, and how it measures and reports model accuracy to brands and marketplaces that send along data. The company has not revealed an explicit business model; in this category, revenue commonly derives from affiliate/marketplace/SaaS agreements directly with merchants.
Regulators are also focused on AI transparency and ads in commerce. Clear delineation of when results and rankings are produced, what data is being used, and how rankings are established will be critical for consumer trust. If Onton maintains its user growth and conversion lift as it enters apparel and electronics, it will serve as a bellwether for whether neuro-symbolic AI can translate discovery into retail performance that sticks — not just strong demos.
