JD Sports rebuilt its product search so AI agents can shop it. Algolia runs the layer

The retailer runs 4,811 stores in 36 countries under ten fascias, from JD and Size? to Hibbett and Go Outdoors. All of them now sit behind one product data layer.

Agents query the same layer that serves the website

An AI shopping assistant does not browse a category page. It asks a question and expects a structured answer. JD Sports has pointed those requests at the same retrieval layer its own sites use, with the same catalogue and the same ranking rules.

Algolia calls this a governed intelligence layer. In plain terms, one copy of the product data answers both the website and any outside assistant, so the two cannot drift apart.

Algolia reports a 22 per cent rise in search revenue

The figures come from Algolia and JD Sports, not from an outside audit. They cover work since 2024, not the agent layer alone.

Revenue from on-site search rose 22 per cent. Click-through on search results rose 7.65 per cent, and on product listing pages by 73 per cent. Revenue from those listing pages grew 16 per cent. A dynamic re-ranking feature added 4 per cent to add-to-cart and 4 per cent to conversion.

Kristin Matter says manual merchandising went first

Kristin Matter, vice president of digital operations at JD Sports, said replacing a manual search and merchandising process changed how the team works. Algolia chief executive Stephen Lynch framed discovery as a growth engine rather than a website feature.

Neither side gave a contract value or said how many assistants now read the catalogue.

Retailers are splitting into feeds and shopfronts

The pattern repeats across the sector. Instacart hands retailers a white-label AI assistant, and research from Profound found that ChatGPT Shopping leans on product feeds rather than on brand pages.

Shoppers are already ahead of the tooling. The Future Shopper study put AI use at 82 per cent while a third still refuse to let it buy.

Open your own product feed and read one record as a machine would. If the title carries the colour but the attribute field does not, an assistant cannot filter on it, and the item drops out of answers your category page would have shown.