A Search Company Just Bought What It Searches

Picture a shopper typing “linen midi skirt with pockets” into a retailer’s search bar and getting pleated maxi dresses. The retrieval worked exactly as advertised. The catalog underneath it never knew the skirt had pockets. Fast search cannot fix what the inventory does not understand about itself. 

Algolia announced on October 6 that it has acquired Velou, a New York startup that uses multimodal Artificial Intelligence (AI) to extract structured attributes from product descriptions and images. Velou’s six-person team joins Algolia, and its catalog enrichment, retail taxonomy and product graph will run beneath the search, recommendations, personalization and merchandising that Algolia already runs for 18,000 customers. Financial terms were not disclosed. 

To understand why Algolia bought the thing it searches, start with where the industry has been racing: retrieval speed. Algolia handles nearly two trillion searches per year, built on an engine known for returning results in milliseconds. Co-founder Julien Lemoine put the shift plainly in his post: retrieval intelligence tells the platform what shoppers mean; Velou tells it what the product actually is. Those are two different answers, and Algolia bought the second one. 

The under-noticed detail is in the chief executive’s own words. CEO Stephen Lynch said in the release,”Algolia won search on speed. The future of discovery will be won on intelligence.” This is a search company’s chief executive calling the dimension his industry has competed on for twenty years a settled race. Algolia still bought a company that does not do search at all. Its own thesis is that the product data has become worth more than the queries that run against it. 

Read as a sequence, Algolia has been stacking intelligence layers on top of its search engine for years. Keyword and vector retrieval, behavioral signals, recommendations and personalization have all accumulated above the base technology. Velou slots underneath that stack, reshaping the raw data the entire pyramid depends on. The acquisition does not add a feature at the top. It changes an input at the bottom. 

This is not an isolated bet. AI agents are increasingly the ones submitting queries to retail catalogs, not just human shoppers typing into search boxes. An agent asking a plain-English question about a product cannot work with a thin product record. Velou CEO Gavin Hewitt said in the press release, “…the race that matters now is contextual and relevant product understanding.” That is a description of what agents need, not what human searchers have been willing to settle for. 

What does this mean for a team with no retail catalog and no plans to buy an enrichment layer? Any searchable asset you own—product listings, documentation, help content, services pages—is being queried by systems that read structured attributes better than they read prose. The gap between what your content says and what a machine can extract from it has just become more expensive to leave alone. 

There is a second signal worth reading in Algolia’s own description of integration. The company says Application Programming Interfaces (APIs), software development kits and the integrations its customers already use will not change. What changes is the depth of intelligence underneath them. A vendor can rewrite the engine without asking the customer to do anything. The flip side is that the customer cannot easily see what has changed either, which places a lot of trust in the vendor’s claim about how different the new intelligence actually is. 

One caution before treating this as the end of search competition. Algolia has not published performance data showing the integrated product improves outcomes against its existing stack, and Velou’s own customer count, more than 50 enterprise brands per Algolia, is modest at the scale Algolia operates. The acquisition proves conviction; it does not yet prove the combined system performs. 

The practical move is to audit your own content the way a machine reads it. Pull your highest-traffic product listing or help article and extract only the attributes an outside system would reliably recognize. If what comes back is thin, that is what an AI agent will see when it queries you on behalf of a customer. 

Algolia built a career on how fast its search could return a result. It just paid for how well that result would understand what came back.