The most important merchandising work on your site this year will never be seen by a person. Agents are doing the browsing. They read attributes, not hero images.
Shopify reported 7x growth in AI-driven traffic since January 2025. Sephora expanded a ChatGPT shopping pilot in March 2026. The browsing layer moved, and it moved fast enough that your last redesign is already answering a question nobody is asking.
Merchandising for agents looks like this. A shopper asks an assistant for a fragrance-free moisturizer under $40 that plays nicely with retinol. The assistant never opens your site. It queries structured data. If your product records carry ingredient flags, skin type, format, size, and compatibility as real fields, you are a candidate. If that information lives in a lifestyle paragraph and a lookbook image, you are invisible.
The work is unglamorous, which is why the lane is still open. Someone has to define the attribute schema for your category, decide which 30 fields actually matter, and fill them across every SKU. Someone has to rewrite product copy so it answers questions instead of setting a mood. On a catalog of 2,000 SKUs, that is months of enrichment, not a sprint. It is a merchandising hire, not a design engagement.
The honest catch is measurement. You will not get a clean dashboard for this. Agent traffic arrives thin on referrer data, some flows convert off your storefront entirely, and you cannot A/B test your way to the right schema. You are spending on a thesis, and you will feel that in the budget conversation. Plan for a 2 quarter horizon before anyone can prove it worked.
Search rewarded the brand that described itself best. Agents reward the brand that describes itself most completely. Those are different jobs, and only 1 of them has a design phase.
