Walmart opens Sparky to Meta Muse and other AI agents: How eCommerce teams can prepare

On October 21, Walmart announced how outside AI agents will shop its store. The Walmart Agent Gateway gives agents a direct connection. Agent Profiles let shoppers control what each agent can spend, and a new set of terms sets the rules agents must follow.

Reuters reports that the Gateway will connect assistants including Meta’s Muse, ChatGPT and Gemini, and Walmart says agents on it connect through Sparky. An agent using the Gateway will read your product data through Walmart’s own assistant, so that data has to hold up for every agent that connects.

Muse has grown quickly since Meta launched it on September 8. It passed 5 million US downloads in its first 22 days, according to Sensor Tower, and more than 10 million users within six weeks, Mark Zuckerberg said on Meta’s third-quarter earnings call on the same day as Walmart’s announcement. Shopping is its most common use, with more than a quarter of weekly active users shopping through it and an average purchase above $200, a figure that includes travel. 

Agent shopping is moving from experiment to habit, and the largest US retailer is now building the infrastructure for it.

What are Walmart’s new tools for AI agents?

Walmart’s new tools for AI agents are the Walmart Agent Gateway, Agent Profiles and a set of terms agents must follow. Agents can already shop Walmart.com through a web browser, clicking around pages built for people. Walmart says that is slow and error-prone, and these tools are its fix.

The Walmart Agent Gateway: This is a direct connection for agents to Walmart and Sam’s Club in the US, built on MCP, the open standard that lets AI systems plug into outside tools. According to Walmart’s beta page, agents connect through Sparky and get up-to-date product and fulfillment data straight from the source. For now it is invitation-only, for companies and developers.

Agent Profiles: Arriving in November, these give each agent its own profile inside a shopper’s Walmart account. Shoppers can approve every order or let the agent spend up to a limit, such as $50, $100 or $200. Walmart compares it to adding an authorized user to a credit card.

Terms for AI agents: Agents need the shopper’s approval by default, never see card details, and can use only the information a task needs. Any agent that breaks the rules will be blocked.

How will AI agents shop through Sparky?

AI agents on the Gateway will shop through Sparky, so they see Walmart the way Sparky sees it. Whatever Sparky knows about your product is what an outside agent will work from.

Our view is that Sparky is becoming the shelf every agent shops at Walmart: optimize for Sparky, and you are helping optimize for every agent that comes through the Gateway.

Sparky also reads beyond Walmart’s catalog, with brands’ own websites making up 21.6% of its citations in August 2026. Walmart hasn’t said how the Gateway ranks products, so how much of Sparky’s own judgment carries over is still an open question.

Walmart also keeps the household’s memory. The Gateway helps an agent find a customer’s usual items without passing their purchase history to the agent, and Walmart’s terms limit agents to the information a task needs. The brand in a shopper’s Walmart history is the one the agent gets pointed back to, which makes the first Walmart purchase the one to win.

Will Walmart ads reach AI agents?

Walmart has not yet said whether its ads will reach AI agents, or how products are ranked for them. It has been testing ads inside Sparky this year, Adweek reported, so the answer matters to any brand planning its 2027 Walmart Connect budget.

The agent side is clearer. Meta says Muse carries no ads and keeps conversations apart from its ad systems, so any paid influence over what Muse buys at Walmart would have to come from Walmart.

How can brands prepare for AI agents and Sparky?

Brands can prepare by starting with their Walmart data, because that is what every agent on the Gateway will read. Walmart already grades listings on the signals agents will use. Its Listing Quality score measures content quality, price competitiveness, shipping, published and in-stock status, and ratings and reviews.

The Gateway passes product and fulfillment data from those same systems to outside agents. Walmart’s published search research, which we unpacked in our Sparky guide, shows how its models judge relevance. Each step below applies one of the 5Cs of Agentic Commerce to that picture.

1. Fill every attribute, recommended ones included (Completeness)

Walmart’s relevance models read a product’s attributes together with its title, adding values such as color to the model’s input. Through the Gateway, those attributes reach outside agents as structured data, so a blank field is a filter your product cannot pass when a shopper asks for something specific, such as fragrance-free, gluten-free or a particular size.

Start with Walmart’s Listing Quality dashboard, which scores every item and suggests fixes, and fill recommended attributes as well as required ones. Then check the fields an agent needs to answer a full request: count and pack size, dimensions, ingredients or materials, compatibility, and age range where it applies.

2. Write for the long, specific request (Context)

Walmart built its semantic search for “tail” queries, the long and specific searches where shoppers describe a need, and for words with more than one meaning, the way “apple” can mean a fruit or a laptop brand. A request handed to an agent reads like a tail query by default, because the shopper describes the whole job in one go: who it is for, the constraint, the budget and the deadline.

Map the jobs your product does and the words shoppers use for them, then make sure your title, key features and description answer those use cases directly. Connect each attribute to the need it serves, so the listing explains what the product is for as well as what it is.

3. Treat your Walmart listing as a quarter of the job (Citations)

Sparky builds its recommendations from far more than the Walmart listing. In Q2 2026, Azoma’s analysis of millions of shopping agent citations found Sparky drew 36% of its sources from earned media, 30% from brand websites, 27% from retailer content, 5% from user-generated content and 2% from affiliate sites.

Align claims, specs and use cases between your brand site and your Walmart listing, so an agent finds one consistent account of the product. Then invest in the earned coverage that makes up the largest share of what Sparky reads.

4. Find and fix what agents get wrong about your products (Correctness)

Run the questions shoppers ask about your top SKUs through Sparky, record every factual claim it makes, and check each one against your product source of truth. Trace each error to the page it came from and fix it there, because through the Gateway, a mistake in your Walmart data reaches every connected agent at once. Keep price, pack size and availability consistent across Walmart, your Shopify store and other retailers too, since agents repeat what they find.

5. Test where agents buy, and date every change (Customer Acquisition)

In the US, ask Muse to buy your three best sellers. It asks for approval before it buys, so you can see which product and retailer it picks without placing the order. Run the same requests in ChatGPT and Gemini and note which retailer each buy link points to.

Log every content change with a date, then read it against your Listing Quality score, agent visibility and Walmart sales for that SKU. Walmart reported on its August earnings call that Sparky users spend 40% more per order than other shoppers, so this is a high-value audience worth measuring closely.

Summing up

Meta and Walmart made the same bet: shoppers will hand more of their purchases to agents. Agents on the Gateway will read Walmart’s data through Sparky, buy within limits the shopper sets, and reorder what a household already likes. That favors brands with complete, consistent Walmart data that are already in the basket. We expect other large US retailers to publish their own rules for agents, and the brands that get Walmart right now will be ready for them.

Azoma is built for this. We measure how AI shopping agents retrieve, compare and recommend products across Walmart Sparky, Alexa for Shopping, ChatGPT, Meta AI and Claude, and we fix the sources those answers are built from. 

Get in touch today to see how your catalog reads to Sparky, and to every agent that will shop through it.

FAQs

How can I track how my products show up in Walmart Sparky?
Brands can track how their products show up in Walmart Sparky by running the questions their shoppers ask Sparky every day, then recording which products it recommends and how it describes them. Azoma does this automatically. It re-runs prompts daily and captures the mentions, citations, rank, sentiment, buy links and carousel recommendations in every answer. Pair that with Walmart’s Listing Quality score to see whether a listing change moved the recommendation.

What tools help products get recommended by Sparky during the holidays?
The tools that matter most for Sparky over the holidays are the ones that keep your Walmart data complete and correct while demand peaks. Adobe expects AI traffic to US retail sites to rise 130% this holiday season, and Walmart’s Agent Profiles arrive in November, so some of that traffic may come from agents shopping through Sparky. Use Walmart’s Listing Quality dashboard to close attribute and in-stock gaps before peak, and an agent visibility platform such as Azoma to track holiday prompts daily and catch price or stock errors in Sparky’s answers.

Can Meta Muse shop at Walmart?
Yes. Walmart already lists Meta Muse among its AI agent integrations, and Reuters reports that Walmart’s new Agent Gateway will connect assistants including Muse, ChatGPT and Gemini through Sparky. Shoppers will be able to set spending limits or approve each purchase through Agent Profiles, which launch in November.

How do brands get recommended by Meta Muse?
Brands get recommended by Meta Muse by making the pages and feeds it reads complete and correct. Muse reads product feeds and live merchant pages, including brand.com and retailer listings, and acts on what it finds.

What is agentic commerce optimization?
Agentic Commerce Optimisation (ACO) is the practice of making a brand’s products discoverable, understandable and recommendable by AI shopping agents, including Meta Muse, Walmart Sparky, ChatGPT and Gemini. It overlaps with GEO (generative engine optimization) and AEO (answer engine optimization): those disciplines win the answer, and ACO carries it through to the purchase.