How AI Agents Are Rewriting the Rules Between Brands and Customers

ChatGPT already drives 20% of Walmart's referral traffic—and the transaction layer hasn't even launched yet

Your customer is about to disappear. Not literally, but functionally. Between the moment they decide they need something and the moment they buy it, there’s increasingly an AI agent making decisions on their behalf. And that agent doesn’t care about your brand equity, your Super Bowl ad, or the fifteen years you spent building shelf presence at Target.

This isn’t science fiction. Traffic to US retail sites from generative AI browsers and chat services increased 4,700% year-over-year in July 2025. ChatGPT is already estimated to drive referral traffic equal to roughly 20% of Walmart’s total visits. And these aren’t casual browsers—AI-referred visitors spend 32% more time on site, browse 10% more pages, and have 27% lower bounce rates than traditional visitors.

Welcome to agentic commerce, where the customer relationship you spent decades building is about to get mediated by an algorithm you don’t control.

What Agentic Commerce Actually Means

The term sounds like consultant-speak, but the concept is simple: AI agents that can complete transactions on your behalf with limited human input. You tell ChatGPT “I need a gift for my DIY dad under $100,” and instead of getting a list of search results, the agent browses products, compares options, considers your past preferences and loyalty memberships, and presents a recommendation—or just buys it for you.

OpenAI launched Operator in January 2025, initially partnering with eBay, Etsy, and Instacart. Perplexity rolled out “Buy with Pro,” allowing one-click purchases from select merchants. Amazon is testing Alexa+ features for $19.99/month (free for Prime members) that let Alexa initiate transactions across Amazon, Whole Foods, Ticketmaster, and other partners. Google’s Gemini has integrated shopping functionality directly into its AI-powered search experiences.

The infrastructure is being built right now. And it’s moving faster than most marketers realize.

The Economics That Make This Inevitable

Here’s why agentic commerce isn’t optional for the platforms building it: OpenAI is losing money on its $200/month Pro tier because users are consuming more compute than expected. The company generated $3.7 billion in revenue in 2024, with 75% coming from subscriptions. But only 5% of ChatGPT’s 800 million weekly users pay anything. That math doesn’t work long-term.

Compare that to Facebook, which generates about $5.12 ARPU globally through advertising. Or Google, which monetized search so effectively it became one of the most valuable companies in history. The playbook is obvious: build massive free user base, then monetize through transactions and advertising.

Industry analysts estimate ChatGPT fields 2.5 billion prompts daily. About 2.1% involve “purchasable products”—roughly 53 million shopping queries per day. If we assume a conservative $25 average order value and an 8% transaction fee, that’s a potential $15 billion in annual revenue from agentic commerce alone. And that’s before the transaction layer even launches properly.

Sam Altman recently said he’s “not totally against” advertising in ChatGPT, specifically mentioning transaction revenue and affiliate models. OpenAI hired Fidji Simo—Facebook’s former VP who specialized in monetization—as CEO of Applications in May 2025. These aren’t coincidences. The company is building the foundation for what could become one of the largest ad networks in history.

Why Amazon’s Walls Are Everyone Else’s Opportunity

Amazon has mostly blocked AI shopping integrations, protecting its own ecosystem. That created an opening for Walmart, whose 420 million SKU catalog now dominates what industry analysts call the “agentic digital shelf.” When ChatGPT browses for products, it’s often pulling from Walmart’s inventory because that’s what’s available.

This matters more than you’d think. In an agentic shopping environment, being in the dataset determines whether you exist to customers at all. If an AI agent can’t access your product information, you’re invisible. And right now, the companies building shopping agents are choosing which retailers to partner with based on who’s willing to integrate and share data.

For brands, this creates a new layer of distribution complexity. You’ve spent years optimizing for Amazon’s search algorithm, Google Shopping, and Instagram’s product tags. Now you need to ensure your product data is structured in ways that AI agents can access, understand, and recommend. And you need to figure out whether you’re comfortable paying transaction fees or advertising to those agents to get prioritized placement.

The Three Types of AI Shopping That Are Emerging

Not all agentic commerce works the same way. Three distinct models are taking shape, each with different implications for how brands show up:

Conversational discovery: The user asks open-ended questions, and the AI narrows down options through dialogue. “I need running shoes for overpronation” becomes a back-and-forth that surfaces specific products. This favors brands with clear, structured product information and strong reputations that get mentioned in training data.

Autonomous shopping: The user gives the AI a goal and it completes the entire transaction independently. “Reorder my usual grocery list but swap in organic where available.” This favors brands with loyalty programs, subscription models, and data integrations that let the AI access purchase history.

Assisted transactions: The AI helps compare options and makes recommendations, but the human completes the purchase. This is the current state for most platforms, and it’s where advertising will likely start—sponsored recommendations within the conversation flow.

Each model creates different opportunities for brands to influence the recommendation. The first rewards content marketing and SEO-like optimization for AI training data. The second rewards operational excellence and integration partnerships. The third will eventually become a straightforward ad auction—brands pay to appear in recommendations for relevant queries.

What Advertising Will Look Like (and Why It’s Coming Fast)

Perplexity already launched sponsored answer placements in late 2024. Users ask a question, get an answer, and see “related questions” that are paid placements. It’s subtle, clearly labeled, and exactly what you’d expect an LLM ad format to look like. You.com integrated sponsored links. Baidu’s Ernie Bot is experimenting with conversational advertising enhancements.

Every major LLM will likely have an ad platform within 18 months. The formats will include sponsored answers (brands pay to appear in relevant responses), contextual placements (ads matched to conversation topics), and affiliate links embedded in recommendations. Unlike banner ads or search results, these will be woven directly into the conversational output.

The ad tech infrastructure is already being built. IAB Tech Lab launched an AI Content Monetization Protocols working group in August 2025 to create standards for how publishers and brands can monetize in AI-driven experiences. Mastercard announced Agent Pay technology in April, allowing verified AI shopping agents to make transactions on behalf of consumers. The pieces are coming together quickly.

For brands, this means the marketing playbook is about to expand again. You already buy search ads, social ads, display ads, and retail media. Soon you’ll buy LLM ads—paying to be mentioned in ChatGPT’s responses, prioritized in Perplexity’s recommendations, and featured when Gemini suggests products.

The Trust Problem Nobody’s Solved

Here’s what makes agentic commerce fundamentally different: trust becomes abstract. When you walk into a store, the trust equation is simple—do I trust this retailer, this brand, this product? When an AI agent shops on your behalf, trust gets filtered through layers of automation and institutional frameworks.

Do you trust the AI to make good decisions? Do you trust that its recommendations aren’t biased by hidden advertising? Do you trust that your payment information is secure when the agent completes transactions? Do you trust that the agent actually compared all available options or just the ones in its training data?

These questions don’t have easy answers. And consumers haven’t figured out their comfort level yet. A recent survey found that 92% of people who’ve used AI for shopping said it enhanced their experience, and 87% said they’re more likely to use AI for larger purchases. But we’re still in the early adopter phase. Wait until mainstream consumers realize the “helpful assistant” is making money every time it recommends something.

The platforms building agentic commerce are acutely aware of this tension. That’s why they’re moving carefully, testing sponsored placements gradually, and emphasizing transparency and disclosure. They know that if users lose trust in the recommendations, the entire model falls apart. But they also know that advertising is inevitable because the economics demand it.

What Brands Should Actually Do Right Now

First, audit how your products appear in AI-generated responses. Ask ChatGPT, Claude, Perplexity, and Gemini to recommend products in your category. Do you show up? How are you described? Are competitors mentioned instead? This isn’t theoretical research—it’s the beginning of your AI search strategy.

Second, structure your product data for AI consumption. That means clean, detailed product specifications, clear use cases, strong reviews and third-party validation, and content that explains not just what your product is but why someone would choose it. AI agents are pulling from web content to make recommendations, so the information they can find about you matters enormously.

Third, build direct relationships with the platforms creating shopping agents. Shopify integrated with ChatGPT early. That gives Shopify merchants visibility in one of the largest AI platforms in the world. If you’re a major brand or retailer, you need to be having conversations with OpenAI, Google, Meta, and the other companies building agentic commerce infrastructure.

Fourth, prepare for the ad networks that are coming. You don’t need to buy LLM ads today, but you need to understand how they’ll work and how to measure their effectiveness. Traditional attribution models won’t capture conversions that happen entirely within a chat interface. You’ll need new tracking mechanisms, different KPIs, and probably new team members who understand how to optimize for conversational AI rather than search algorithms.

Finally, accept that your direct relationship with customers is getting mediated—again. First it was retailers. Then Amazon. Then Google. Then social platforms. Now it’s AI agents. Each layer adds efficiency and convenience for customers while reducing your control over the relationship. The brands that succeed won’t be the ones fighting this trend. They’ll be the ones figuring out how to be the recommendation that AI agents make.

The Bigger Picture

Agentic commerce is happening whether brands are ready or not. The technology works. The business models are proven. The user behavior is shifting. And the economic incentives are overwhelming.

Within five years, a meaningful percentage of online shopping will flow through AI agents rather than traditional e-commerce sites. Those agents will be monetized through advertising, transaction fees, and affiliate revenue. The platforms that own those agents will become some of the most powerful gatekeepers in commerce—potentially more influential than Google or Amazon because they’ll control not just discovery but the entire decision-making process.

For marketers, this isn’t a crisis. It’s a transition. You’ve navigated these before. You figured out search marketing when Google launched AdWords. You learned social advertising when Facebook opened its platform. You adapted to retail media when Amazon started selling ads. You’ll figure out agentic commerce too.

The question is whether you’ll do it proactively or reactively. Because the brands that understand this shift early and invest in being AI-findable will have an advantage that compounds over time. And the ones that wait until agentic commerce is fully formed will be playing catch-up in a market where the rules have already been written by someone else.

The shopping middleman you didn’t know you needed is already here. Your job is to make sure it recommends you.