Instacart Prepares Ads for Cart Assistant Conversations
Instacart is preparing to test advertising formats inside its AI-powered Cart Assistant in the second half of 2026, according to Adweek. The company has outlined sponsored recipe and occasion placements, deal surfacing, and product-exploration formats for conversational grocery shopping, PYMNTS reported. Cart Assistant uses historical order data and live store inventory, while Instacart reports that orders placed through it exceed its $115 average order value.
Instacart is preparing to test ads within its AI-powered Cart Assistant during the second half of 2026, according to Adweek. The proposed formats would put sponsored content into grocery-planning conversations, extending Instacart's retail-media business into an interface where shoppers ask for recipes, compare products, and build carts.
Cart Assistant has rolled out to millions of US users, and Instacart announced a broader US and Canada rollout in June, Adweek reported. The assistant draws on data from more than 1.6 billion lifetime orders and real-time inventory from nearly 100,000 stores across North America, according to Adweek and PYMNTS.
PYMNTS reported that Cart Assistant orders are larger on average than Instacart's reported $115 average order value. That comparison does not establish that the assistant alone caused larger baskets, but it provides the commercial context for testing ad placements within the experience.
Three proposed conversational formats
Reporting by Adweek and PYMNTS describes three ad concepts tailored to Cart Assistant interactions:
- •Sponsored recipes and occasions, intended to surface brands when a shopper is planning a meal or event, such as a barbecue.
- •Deal surfacing, which would bring offers into a shopper's conversation.
- •Product exploration, intended to place brands in product-comparison or decision-making interactions.
These differ from conventional sponsored search because the input can be a multi-step request rather than a product keyword. In comparable retail-media deployments, conversational placement raises technical questions around relevance ranking, disclosure, measurement, and whether paid recommendations preserve the usefulness of the assistant's response. The available reporting does not specify Instacart's ranking rules, auction mechanics, labeling approach, or measurement methodology for the planned pilots.
Retail media expansion
Instacart reported more than $1 billion in advertising and other revenue for 2025, with reach across nearly 100,000 stores and more than 2,200 retail banners, PYMNTS reported. Separately, Instacart reported $286 million in first-quarter advertising revenue, up 16% year over year, according to ContentGrip.
The company has also introduced an in-app Immersive Feed for shoppable short-form video and an Ads Studio offering, Marketing Dive reported. Those initiatives place the Cart Assistant test within a wider retail-media expansion from sponsored listings toward discovery-oriented formats.
For ML and data teams building commerce assistants, this pilot is a useful case to watch because it combines recommendation, retrieval of live inventory, and advertising in one workflow. Industry experience with similar systems suggests that measurement will need to distinguish assisted conversion from organic demand, while product-quality evaluation will need to account for sponsored-result relevance rather than click-through rate alone.
Key Points
- 1Instacart is preparing conversational ad pilots, bringing sponsored recipes, deals, and product exploration into its Cart Assistant workflow.
- 2Cart Assistant combines 1.6 billion historical orders with live inventory, making relevance and availability central to conversational recommendations.
- 3Comparable commerce assistants require measurement that separates incremental ad outcomes from organic purchase intent and recommendation quality.
Scoring Rationale
The reported ad pilots are a notable commercialization test for conversational AI in a large grocery-commerce environment. The story is relevant to practitioners working on recommendation, retrieval, ad ranking, and measurement, though the formats remain in testing rather than broad deployment.
Sources
Public references used for this report.
Practice with real Food Delivery data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Food Delivery problems
