The retail media playbook has had a fairly stable shape for five years: Amazon owns search, Walmart's Connect ecosystem has been the credible second house, and everybody else is fighting for the long tail. That shape is now changing fast.

Digiday reported overnight — and Walmart Connect leadership confirmed during a Monday morning industry briefing — that Sparky, Walmart's agentic shopping assistant, will begin carrying sponsored placements as a formal ad inventory line in 2026. Two details from the announcement matter more than the headline.

First, Sparky will be portable. Walmart's EVP of AI Acceleration, Hari Vasudev, told industry partners that "the Sparky experience will travel directly into ChatGPT and Gemini and anybody else that we ever integrate with." That mirrors what the company has signaled since Walmart's Universal Commerce Protocol announcement at NRF in January, but it's the first time Walmart has explicitly framed Sparky as a destination-agnostic agent rather than a Walmart.com feature.

Second, the basket data is real. Walmart disclosed in its 2026 annual report this past month that Sparky drives "about 35% higher baskets" for users who interact with it. Whatever you think of the agentic-commerce hype cycle, that is a hard, measurable number — and one Amazon's Rufus has not publicly matched.

The competitive picture

Walmart isn't moving in a vacuum. Carbon6's running scorecard on the agentic-AI assistants tracks the divergence:

  • Amazon Rufus has been carrying ads since late 2024. It's tightly integrated to Amazon's product graph and sells inside Amazon's apps and site.
  • Walmart Sparky went GA across walmart.com last fall. As of today, it's the first agentic assistant credibly trying to leave its native surface and carry the merchant's intent into ChatGPT or Gemini.

That distinction is the whole strategic ballgame. As PYMNTS framed it this morning, the war is no longer for the buy button — it's for what they call "the decision layer," the layer above checkout where the consumer's intent gets resolved before any specific store is selected. Whoever owns that layer for grocery and household essentials owns retail's most defensible long-term margin pool.

What "ads in Sparky" actually means

The mechanics of the new Sparky ad inventory haven't been fully detailed, but Mirakl's retail media outlook and Digiday's reporting suggest three product lines:

  1. Sponsored answers. When a shopper asks Sparky for "a quick weeknight dinner for four" or "what should I get my niece for graduation," brands can pay for placement in the recommendation set, similar to Sponsored Products on Amazon — but inside an agent's natural-language output.
  2. Sponsored substitutes. When Sparky recommends a substitute for an out-of-stock item, brands can bid for that substitution slot. This is a category Amazon has not formally productized.
  3. Co-branded creative. Generative AI creative tooling that lets brand partners produce Walmart-Connect-compliant ad assets at the SKU level. Mirakl's 2026 trends report flagged this as the largest category of new spend going into Q3.

The CPG implications

For CPG brand teams, the most important consequence is that media buying just got harder. Targeting in an agentic system is fundamentally different from keyword bidding on a search results page. You're not optimizing for a query; you're optimizing for a multi-turn conversation that may surface your product in turn three when the consumer never typed your category.

eMarketer's FAQ on agentic commerce projects that AI platforms will account for 1.5% of total U.S. retail e-commerce sales in 2026 — about $20.57 billion — quadrupling 2025's level. That is still small in absolute dollars, but it's the fastest-growing percentage of any retail-media line by an order of magnitude.

Two practical implications brands should be working through this week:

  • Audit your product feed. If your data isn't structured for agent retrieval — clear category, attribute, occasion, and substitution metadata — you're invisible to Sparky regardless of how much you spend on existing Walmart Connect.
  • Re-think the "always-on" budget split. Brand teams that have allocated 70% of retail media to Amazon Sponsored Products as a default should be pressure-testing that split against where agentic queries will route in 12 months.

Why this matters for the rest of retail

The Sparky-inside-ChatGPT pitch is also a warning shot to every retailer below the top two. Modern Retail's reporting on Criteo's Q1 earnings flagged the same dynamic: smaller retail-media networks risk getting structurally disintermediated when consumer intent now resolves at the agent layer rather than the retailer search bar. Target, Costco, Kroger, and Best Buy all have credible retail-media businesses today. None of them have yet articulated how their agent strategy keeps that revenue line growing in a world where Sparky, Rufus, and a future OpenAI-native shopping agent are all fighting for the same upstream intent.

Walmart's move today says that's the next 18 months of strategic work for everyone in retail media, not just the top two.