An AI Assistant on Your Site and Being Recommended in ChatGPT Are Two Jobs

A board member forwards the AWS news. A vendor says the on-site assistant covers "AI shopping". The question a Head of Ecommerce has to answer is simple: if we buy an AI shopping assistant for our storefront, does that also take care of whether ChatGPT and Google AI Mode recommend us? It does not, and the two jobs deserve two budgets.

What AWS is selling

AWS launched the Agentic Shopping Assistant on May 27, 2026. CNBC reports that Amazon took the architecture, starter code and learnings from its own Alexa for Shopping and packaged them for other retailers, with launches "in as little as 60 days." Kate Spade, owned by Tapestry, is the first named customer.

The Kate Spade gift concierge went live in April after roughly two and a half months of testing, built on Anthropic's Claude Haiku 4.5 through Amazon Bedrock, according to GeekWire's coverage. AWS says retailers keep control of their customer data, product catalogs and business rules. SiliconANGLE adds that the service runs on Bedrock AgentCore and OpenSearch.

In plain terms: a conversational agent that lives on your site, reads your catalog, talks to shoppers who have already arrived and hands them to checkout. For a mid-market retailer without a machine learning team, that is a real product with a real use.

The gate that decides whether the shopper arrives

Adobe's data, reported by Adweek in late May, puts traffic to retail sites from AI sources up almost 400% year over year in the first quarter of 2026. That growth comes from ChatGPT, Google AI Mode, Perplexity, Gemini and Claude, not from retailer storefronts.

When a shopper asks ChatGPT for a gift for a friend who has a new baby, the answer is assembled before any retailer gets to greet that shopper. The brands that appear are chosen from product information the assistant can read: product pages, Merchant Center feeds, structured data and what third parties say about the product.

An on-site assistant sits downstream of that choice. The Kate Spade concierge can turn intent into an order once the shopper is on katespade.com. It has no influence on whether ChatGPT mentioned Kate Spade when the shopper first asked. Two problems, two remedies.

Two jobs, side by side

What an on-site shopping assistant does:

• Answers questions and recommends products for shoppers already on your site
• Reads your own catalog and your own business rules
• Improves conversion and gives you first-party conversation data
• Shows up on the vendor's dashboard as sessions, chats and assisted orders

What AI visibility work does:

• Checks whether your products appear when shoppers ask ChatGPT, Google AI Mode and Perplexity what to buy
• Fixes the product information those assistants read, on the page and in the feed
• Tracks mentions, product cards and AI referred sessions before and after each fix
• Covers surfaces you do not own and cannot configure

Buy the first and assume the second is handled, and the likely result in the back half of the year is better conversion on your own site while your share of AI answers stays flat. Both numbers matter. Only one is on the AWS dashboard.

The dependency question

The company offering the agent is also your largest competitor in product search and the marketplace that already ingests your catalog. CNBC notes that offering the tool through AWS could reassure retailers wary of sharing data with Amazon, and AWS says retailers keep ownership of their data.

That is a contract question, not a reason to refuse the deal. It is a good reason to keep the work that decides whether ChatGPT or Gemini mentions your brand outside the same vendor relationship. A single vendor is acceptable for a conversion layer. It is a riskier choice for a discovery layer where a competitor runs the surface.

Three decisions for the next 30 days

1. Separate the two budgets before any vendor conversation

One line for an agent on your storefront. One line for being recommended inside assistants you do not own. Decide which is more urgent for your category and your acquisition plan. The first matters once shoppers arrive. The second decides whether they arrive at all.

2. Find out whether shoppers see you when they ask

Write ten to twenty questions the way your customers phrase them, with budget, recipient, material or use case, and ask ChatGPT and Google AI Mode. Record which brands and products come back, and whether yours appears as a mention, a link or a product card. Then tag AI referrers as their own channel so you can see the sessions. We covered that setup in how to measure AI referral traffic in GA4 and Shopify.

3. Fix product data, because both gates read it

The on-site assistant reads your catalog. So do ChatGPT's shopping results and Google's Shopping Graph. Thin descriptions, missing attributes and stale prices produce a polite but wrong answer from your own assistant and an absent product card in ChatGPT. Complete, accurate product content is the prerequisite for both investments. A free AI readiness report on one representative product shows where the gaps are.

Frequently asked questions

Is the AWS assistant a competitor to Salesforce Agentforce or Shopify Sidekick?

Functionally yes. All three are first-party assistants that live on the retailer's storefront and work on conversion after the shopper arrives. The AWS version carries the Alexa for Shopping lineage and runs on Bedrock. None of them change how your products appear inside ChatGPT, Gemini or Google AI Mode.

Does an on-site assistant affect whether ChatGPT shows my product as a card?

No. Whether a product appears in ChatGPT, and whether it appears as a buyable card rather than a passing mention, depends on the product information ChatGPT can read and on the feed sources it draws from. We explain those two decisions in how ChatGPT Shopping picks products and then picks merchants.

If we deploy the AWS assistant, do we still need AI visibility work?

Yes, if new customer acquisition matters to your growth plan. The assistant improves what happens after a shopper lands. AI visibility work improves whether the shopper lands. With AI sourced traffic growing at the rate Adobe reports, the second question is at least as urgent as the first.

Does using the AWS assistant put our catalog inside Amazon's infrastructure?

Model calls and conversation logs run through AWS systems. AWS says retailers retain control of their customer data, catalogs and business rules. Ownership and operating location are different questions, so retailers in categories where Amazon competes directly should structure the contract with that in mind.

The AWS launch is real news about exactly one of the two gates that decide whether a brand wins in AI shopping. Treat it that way, and give the other gate its own owner and budget.

That second gate is the work Paz does for lean ecommerce teams. A Paz expert checks whether your products show up when shoppers ask ChatGPT and Google AI Mode, finds the gaps in your product data that keep you out of the answer, and proposes the fixes with the evidence behind each one. You approve, and the platform publishes the changes and tracks the before and after. Talk to us.