Mars tests how AI shopping agents find snack brands
Mars is probing LLM product visibility and reorganizing commerce teams as brands prepare for AI shopping without proven sales results.
Mars Snacking is testing how its brands appear in AI-generated shopping recommendations while reorganizing teams around a more connected approach to commerce. The new details offer an unusually practical view of what preparing for agentic shopping looks like inside a large consumer-goods business.
Neil Reynolds, Mars's global chief customer officer, described exploratory partnerships with emerging technology companies to understand how products surface in large language models and which information sources influence those results. The work is still about establishing a baseline, not announcing that AI assistants are already buying groceries for Mars customers.
The effort builds on Mars's October research into the future of shopping, developed with Kantar and Synthesis. That earlier work forecast two parallel developments: more routine purchases handled automatically, and physical stores becoming more experiential. The more recent disclosure is about the organizational and measurement steps Mars is taking in response.
Table of contents
Jump to each section:
- Why AI discovery is becoming a shelf-space question
- One Demand squads connect brand, data and retail
- The physical store is not disappearing from the strategy
- What marketers should know about AI shopping
Why AI discovery is becoming a shelf-space question
Consumer brands have long competed for search visibility, retail listings and prominent positions in-store. AI shopping assistants introduce another layer between a shopper's request and the products that receive consideration. Mars is starting by asking what evidence explains whether its brands appear in those answers.
Reynolds framed a central question as “how do we make sure our brands are discoverable by the LLMs”. His answer was not a finished optimization system. The company is working with emerging technology partners to learn how its products currently show up, and where that information comes from.
An AI shopping assistant may not have an aisle, but it still shapes which products enter consideration.
For marketers, the distinction between being mentioned and being purchased matters. A brand could become visible to an assistant without winning a transaction. Mars is therefore treating discoverability and conversion as connected but separate problems, rather than assuming improved AI visibility automatically leads to growth.
That reasoning also makes source quality more consequential. Product descriptions, retailer catalogues, availability information and other underlying records can influence how a brand is represented, although the interview does not disclose which sources Mars has found to be most influential. Claims about an AI-visibility improvement would require measured evidence that the company has not yet published.
One Demand squads connect brand, data and retail
Mars is also changing who works on the problem. Reynolds described forming One Demand performance squads that bring together brand management, consumer insights, key accounts, category teams and agency partners. The structure is designed to connect the people responsible for consumer demand with those closer to retail execution.
That is a more substantial organizational response than assigning AI search to an SEO specialist. The brand team may understand positioning, the account team may understand retailer requirements, and the data team may identify gaps in product information. An answer generated outside any one of those teams still depends on decisions made across all of them.
The deeper constraint may be coordination rather than a lack of new tools.
Part of the infrastructure predates this interview. In June 2025, Mars appointed Publicis to lead connected commerce and several other marketing functions. Reynolds now describes the partnership as a route to linking media, commerce and relevant data more closely. The agency arrangement is not a new deal, and Mars has not attached a new financial or performance result to its AI-shopping work.
An organizational structure can make experimentation easier to evaluate. It does not prove that shoppers are already letting software make independent buying decisions, or that the new approach outperforms existing retailer marketing.
The physical store is not disappearing from the strategy
A familiar assumption about agentic commerce is that less manual shopping must mean less opportunity for impulse purchases. Mars is preparing for a different possibility: automated replenishment for routine items could coexist with stores that offer discovery, experiences and reasons to try something new.
That distinction is especially important for snacks. A product bought on a predictable weekly cycle presents a different marketing challenge from a spontaneous purchase prompted by an in-store display or an unexpected flavor. Mars is discussing store-of-the-future ideas with major retailers, while considering how technology teams on both sides should work together.
The company has not identified a specific new store format or a measured lift in impulse purchases from this work. These are strategic planning conversations rather than evidence of a completed rollout.
The wider commerce market offers different starting points. Instacart's ChatGPT shopping integration approaches the problem through transaction pathways, while Google's Universal Commerce Protocol addresses technical connections for AI-led commerce. Mars's present focus is earlier in the advertiser's process: understanding brand presence in assistant answers and making its commercial teams capable of responding.
The comparison matters because platform integrations and brand readiness are not the same milestone. Retailers can make automated transactions possible without resolving every product-information or measurement challenge that advertisers face.
What marketers should know about AI shopping
Mars's approach suggests that AI commerce deserves attention before any single assistant becomes a dominant purchasing interface. But marketing teams should distinguish preparation from proven consumer adoption.
Visibility needs a baseline. Mars is starting with how its brands appear in model outputs and which sources are involved. Marketers need comparable evidence before interpreting fluctuations as performance gains.
Commerce data crosses team boundaries. One Demand squads illustrate why brand, category, account and agency decisions may need to be assessed together when discovery shifts across platforms.
Physical retail still has a role. Automated replenishment is not identical to product discovery. The relevance of sampling, merchandising and store experiences depends on the shopping occasion.
Attribution remains unsettled. A mention in an assistant answer is not an order. Any claim about incremental sales will require a clearer connection between recommendation exposure, retail action and the resulting purchase.
AI may change where the consumer first encounters a brand, but it does not remove the need for clear product information, retailer cooperation or credible proof of demand.
That is the more useful lesson from Mars's early work. Rather than describing an inevitable future where all shopping is delegated to agents, the company is showing how marketers can organize around an uncertain one. The competitive question is not only whether a brand appears in an AI response. It is whether the business can understand what that appearance means and act on it.
