Coca-Cola vs PepsiCo: The New Battle for the AI Shelf

For decades, Coca-Cola and PepsiCo fought for inches of refrigerated metal. Then search changed how people found a drink before they reached the store. Now the shelf is becoming harder to see.

A shopper can ask an AI what to buy for a barbecue and receive a shortlist before visiting a retailer, searching Google, or looking for a brand directly. The commercial question is shifting from where are you stocked? to does AI recommend you?

This Insight uses Coca-Cola and PepsiCo as recognizable anchors for a broader CPG shift. It is not a FutureFox ranking. We did not score either company for AI Readiness, and we are not claiming one is more recommendable than the other. The title names the rivalry because every CPG leader already understands that fight. The new fight is less visible.

What is the AI Shelf?

The AI Shelf is FutureFox's name for the recommendation layer where AI systems decide which brands and products are worth presenting in response to a commercial question. It is not a literal shelf. It is a conceptual layer between consumer intent and brand consideration. We are not presenting the phrase as an established industry term. It is an editorial framework, inspired by a wider 2026 conversation already underway.

Google Cloud describes an invisible shelf where AI agents research, find, and in some cases purchase products. Deloitte writes about an algorithmic shelf where products can be ranked, substituted, or excluded without a human rejection showing up in existing KPIs. FutureFox's contribution is more current: most CPG brands are not yet competing for fully autonomous purchase. They are competing to be named when a shopper asks what to buy.

A traditional results page can hold ten brands. A recommendation often holds two or three, plus a reason. NIQ and Kearney put the mechanic plainly: instead of a hundred results to parse, the system may return a shortlist with an explanation. If you are not on that shortlist, you were never in the running.

How the shelf evolved

Physical shelf: retailers choose what to stock. Scale, trade spend, packaging, and relationships still decide much of that. Coca-Cola and PepsiCo spent a century becoming fluent here. Digital shelf: ecommerce search, marketplace ranking, retail media, and Google results surface a different assortment. Amazon and Walmart made this layer unavoidable. AI Shelf: a shopper asks a constrained question, and the system recommends. That is the layer this article is about.

Agentic commerce sits further out: AI may later help select and purchase. Google Cloud, NIQ, Kearney, and Deloitte all treat this as a serious 2026 planning issue. It is not yet mainstream behavior. Gartner's January 2026 U.S. survey found only 11% of consumers willing to let AI make purchase decisions, even in lower-stakes categories such as personal care and household supplies. The AI Shopping Trust Gap 2026 reviews that evidence in full. Plan for the destination. Compete now for the recommendation.

How is AI changing CPG product discovery?

CPG has historically won through awareness, distribution, shelf placement, advertising, retailer relationships, packaging, promotions, and repeat purchase. If a shopper thought "cola," two names arrived unbidden. AI changes part of that equation. A consumer asking for best, cheapest, healthiest, zero sugar, family friendly, best for a party, best for sensitive skin, or best after workout is no longer retrieving the most famous brand by default. The prompt introduces attributes. The system can evaluate them.

Published 2026 evidence shows the behavior is already commercial, even if definitions of "using AI" vary. In March 2026, NIQ and Kearney reported that 74% of shoppers are using AI for some form of product discovery, 54% for research, and 20% directly for shopping. A separate NIQ study with the Digital Shelf Institute found 68% of consumers using generative AI at least monthly, with roughly a third or more using it specifically to research products or summarize reviews. The gap between those figures is useful. Discovery is moving. Delegation is slower.

Adobe shows the traffic consequence. In Q1 2026, AI-referred visits to U.S. retail sites grew 393% year over year. By May, Adobe reported 1,324% growth since October 2024, converting 42% better than non-AI channels in March. Machine-readability has not kept up: product pages scored 66% on Adobe's AI Content Visibility Checker. For Unilever, P&G, Nestlé, Mondelez, or Mars, that is a portfolio problem. Fame still matters. Machine-readable evidence is becoming the price of remaining in the answer.

Coca-Cola vs PepsiCo, without a scorecard

What happens when two globally recognized brands enter an environment where the shopper no longer asks "which brand do you know?" and instead asks "which brand is best for this situation?" The prompts below are hypothetical. They are not FutureFox measurements of either company.

• "Which zero-sugar cola should I buy?"
• "What is the best drink for a barbecue?"
• "Which sports drink is best after a workout?"
• "Which beverage brand has the strongest zero-sugar range?"
• "What's a good family-friendly snack brand?"

Each question adds intent, attributes, constraints, context, and comparison. Zero sugar is not the same job as barbecue. A sports drink after exercise may pull Gatorade, owned by PepsiCo, or Powerade, owned by Coca-Cola, or a challenger such as Celsius. A family snack prompt may leave cola entirely and land in Frito-Lay, Mondelez, or Nestlé. The rivalry does not disappear. It is re-cut by need state.

How AI search recommends brands is less about heritage and more about whether the system can resolve a product against a job. How ChatGPT recommends products depends on entity clarity, structured evidence, and third-party corroboration, not on who won last year's Super Bowl ad.

Can challenger brands outperform famous brands in AI recommendations?

They can, in specific recommendation contexts. They do not automatically win. NIQ retail measurement found that established niche brands increased U.S. market share by 1.5 percentage points from 2022 to 2025, while large and mid-size national brands declined by 2.1 points. NIQ and Kearney argue that AI-led discovery is accelerating that pattern in pet care, personal care, and health and wellness. Kearney's Katherine Black: AI systems prioritize clarity and relevance, and brands that make products legible with structured data, defined need states, and credible signals are better positioned to surface.

A challenger with a sharper attribute story can be easier to recommend than a famous brand with a foggy one. Clearer attributes, stronger review consensus, better comparison content, and tighter use-case positioning all help a model defend a shortlist. Olipop, Poppi, and Celsius are useful as illustrations because each sells a need state as much as a flavor. That does not mean they are winning ChatGPT. It means their commercial story is already written in the grammar AI uses: *what is this for?*

L'Oréal faces the same mechanic in beauty. "Best for sensitive skin" is an attribute match, not a logo match. In February 2026, Unilever announced a five-year partnership with Google Cloud framed around brand discovery as journeys shift toward conversational and agentic experiences. That signals how seriously large CPG now takes the layer. It is not proof of recommendation share. GEO and AI Search Optimization matter here because a constrained prompt can reduce the advantage of pure familiarity. Scale still helps everywhere else. On the AI Shelf, evidence gets a vote.

Why does product data matter for AI Search?

Human brand recognition is not the same as machine understanding. A famous CPG brand can still be illegible to the systems mediating discovery.

A machine-readable CPG brand needs clear attributes, structured product data, accurate information, consistent identity, strong reviews, authoritative mentions, comparison content, availability, trustworthy claims, and fresh data. Google Cloud's CPG guidance is blunt: treat product data as the new packaging. If a product uses sustainable packaging, an agent searching for verified sustainable packaging will not find it unless that information is structured and tagged.

This is where SEO, Answer Engine Optimization, GEO, Entity SEO, and AI Readiness stop being separate slogans. They are how a brand becomes recommendable. None of them replace distribution. All of them decide whether the AI Shelf can stock you. NIQ's June 2026 Product Intelligence launch exists because fragmented attributes and unresolved product identity are now discovery issues. Adobe's product-page readability gap is the public-web version of the same problem.

The Evidence Graph behind the AI Shelf

FutureFox describes the network of proof a model can assemble as an Evidence Graph. For a CPG brand, that graph is rarely the brand site alone: product pages, structured information, retailer listings, reviews, editorial comparisons, creator discussion, and other authoritative sources. The important question is not "did we publish a campaign?" It is: does the web contain enough consistent evidence for AI to confidently recommend the brand?

NIQ notes that models fed contradictory claims do not average them into a safe middle. They lose confidence. A Coke Zero claim that differs across a brand page, a grocer, and a review corpus is a recommendation risk. Premium ecommerce teams already see this in ChatGPT, Gemini, and Perplexity visibility. CPG adds a harder version: more SKUs, more retailers, more claim regulation, and less D2C control. If you do not sell direct, SEO on the brand site is necessary and insufficient. The graph has to hold at Walmart, Amazon, Tesco, and the review layer too.

The invisible shelf is not the end of search

Search is not dead. Retailer search is not dead. Physical shelves are not dead. Brand websites are not dead. AI Search is another discovery layer. Google Cloud still puts brand equity, physical availability, and packaging in the room. Adobe's AI traffic is valuable because it still lands on retail sites. The strategic shift is interaction, not replacement.

The emerging journey is: consumer question, AI recommendation, brand consideration, verification, retailer or brand site, then purchase. The AI Shelf gets the brand into consideration. The Trust Gap determines whether the consumer believes the recommendation. Gartner found more openness to tools that narrow choices than to tools that decide: 31% were willing to let AI narrow household-supplies choices, 28% for personal electronics, and 11% to let AI execute the purchase. Help is welcome. Control is not negotiable.

How should CPG brands prepare for AI Search?

The work is not a rebrand. It is making existing assets legible to the systems now mediating discovery.

1. Make product truth machine-readable

Attributes, claims, ingredients, certifications, variants, and use cases should be structured, consistent, and current. This is how a product gets onto the AI Shelf.

2. Strengthen entity clarity

Parent companies, house brands, and sub-brands confuse machines. Coca-Cola, Sprite, and Minute Maid are related and not interchangeable. PepsiCo, Gatorade, and Lay's the same.

3. Build evidence beyond the brand site

Retailer content, reviews, comparisons, and independent coverage are part of the Evidence Graph. A beautiful owned site with a thin off-site graph still looks sparse to a recommender.

4. Monitor commercial AI prompts

Track the questions that map to revenue: zero sugar, sensitive skin, after workout, family snacks, party drinks. Prompt panels are category research now, not a stunt.

5. Measure recommendation visibility

Share of shelf and share of search are incomplete. NIQ describes share of conversation and share of recommendation as the emerging KPIs. Know whether you are named, omitted, or substituted on the prompts that matter.

6. Protect consistency across retailers

Contradictory titles, claims, and nutrition facts across Amazon, Walmart, and grocers lower confidence. Digital shelf execution is now upper-funnel for AI.

7. Prepare for agentic commerce without overreacting

Build data foundations that future agents can use. Do not budget as if most shoppers will let AI complete the basket this year. Gartner's 11% figure should keep that ambition honest. Teams that want a diagnostic starting point can run the AI Readiness Assessment and use the enterprise ecommerce AI Search checklist as an operating backlog. CPG still has to extend the same discipline across retailer feeds.

The FutureFox view

For decades, the shelf was physical. Then it became digital. Now it is becoming conversational. The winning CPG brands will not abandon distribution, search, or brand building. They will make sure those assets are legible to the systems increasingly mediating discovery.

Executive takeaway

Frequently asked questions

Summary

Coca-Cola and PepsiCo made the physical shelf a global spectator sport. Search then taught CPG to fight for rankings and digital facings. The next contest is quieter. When a shopper asks an assistant what to buy, a small number of products get named. FutureFox calls that the AI Shelf.

We did not measure who wins it today. The brands that will are the ones whose product truth, retailer consistency, and evidence graph can survive a constrained question. For decades, CPG brands fought for the shelf. Now they may have to fight for the AI Shelf.

To see how AI currently represents your brand, run the AI Readiness Assessment, read how AI search recommends brands, or contact FutureFox Labs.

Sources and notes

This Insight is a structured reading of published 2026 research, not a proprietary Coca-Cola vs PepsiCo benchmark. FutureFox did not conduct a CPG prompt panel for this article. Usage statistics vary by definition and method; where sources disagree, we report the figures separately. The AI Shelf and Evidence Graph are FutureFox frameworks. Google Cloud's "invisible shelf" and Deloitte's "algorithmic shelf" are cited as adjacent industry language.