Gap has introduced AI-assisted shopping through its Alta Daily and Daydream platforms, with Banana Republic and Old Navy also preparing conversational shopping agents. Three decades after launching its first website, the company is rebuilding the front door of discovery, and the design choices it makes now will determine whether the group appears in AI answers as a recommendation or disappears into an undifferentiated catalog.Gap AI-assisted shopping
What the Company Rolled Out
Alta Daily and Daydream act as recommendation surfaces rather than storefronts. A shopper describes an occasion, a fit preference or a budget, and the system returns a curated set rather than a search results page. The format suits apparel, where the constraint is usually style context rather than a specific SKU.Alta Daily and Daydream platforms
The conversational agents planned for Banana Republic and Old Navy extend the same logic to two very different price positions. Banana Republic needs guidance on fit and fabric at a higher price point, while Old Navy needs speed and size availability at volume, and one agent design rarely serves both well.
Why Apparel Needs a Different Agent
The Query Is a Situation Not a Keyword
Apparel shoppers rarely search for an attribute in isolation. They describe a situation: a fall wedding, a return to office, a trip where one pair of shoes has to work for three outfits. A conversational agent can accept that input, while a keyword engine has to guess which attribute matters.
Fit Is the Return Driver
Most apparel returns trace back to fit, and fit is the hardest attribute to encode. An agent that asks two clarifying questions about height and preferred looseness can reduce returns more than a size chart ever will, because it captures the preference that the chart cannot express.
Catalog Depth Beats Model Size
Recommendation quality depends more on catalog structure than on the underlying model. If each item carries consistent attributes for occasion, silhouette, fabric weight and care, a modest model produces useful answers. If attributes are inconsistent, a stronger model produces confident mistakes.
The Trade-Off Brands Keep Missing
Conversational shopping improves discovery and worsens comparability. A curated set of five items hides the other four hundred, which helps conversion but can obscure price positioning. Retailers have to decide whether the agent optimizes for a confident decision or for a transparent comparison, because the two goals pull in opposite directions.
There is also a measurement problem. When an agent recommends an item the shopper did not search for, standard attribution credits the recommendation surface rather than the merchandising decision behind it. Without new instrumentation, buying teams lose the feedback loop that tells them which attributes drive discovery.
What AI Answers Require From Brand Content
Assistants outside the brand's own properties need extractable facts: materials, measurements, care instructions, price bands, availability. Gap's own agents can work with internal catalog data, but external assistants can only cite what is published. The gap between internal richness and external thinness is where competitors win citations, and platform agents aimed at small businesses widen it by giving every merchant the same surface.Meta Muse AI agent
The practical fix is a public product data layer: one page per item with structured attributes, a stable URL, and an update timestamp. Retailers that publish this layer get cited consistently, while those that keep specifications inside images or PDFs remain invisible to every assistant except their own.
How to Judge Whether It Works
Three tests are enough for a first assessment. Ask the agent for an outfit under a stated budget and check whether the answer respects the constraint. Ask for a fit recommendation for an unusual body type and check whether it asks a clarifying question. Ask about a specific item and check whether the answer matches the published specification.
An agent that passes all three is a usable discovery surface. One that passes only the first is a marketing demo. The difference becomes visible at scale within a single season, because shoppers who receive a wrong fit recommendation do not complain, they simply return the item and do not come back.
