Brands Are Losing Control of Their Product Catalogs. And L'Oréal, Unilever, Mars Just Fixed It.

You spent years building your website. SEO. Product pages. Brand equity. The whole infrastructure designed to put you in front of buyers.

AI agents don't look at any of it.

When someone asks ChatGPT what ski helmet to buy, the agent doesn't visit your site. It pulls product data from Reddit. From outdated affiliate blogs. From wherever it finds something that looks relevant. Your brand gets represented by whatever the agent scrapes together: no control, no accuracy, no brand governance.

The thesis: AI agents don't browse product pages. They synthesize distributed data. Brands that structure and govern that data now will own their categories in agentic commerce. Brands that don't will be represented by whoever got there first.

That's the problem Azoma's Agentic Merchant Protocol (AMP) solves.

On March 12, Azoma launched AMP as the first enterprise infrastructure that gives brands a single system to define, distribute, and govern how their product catalogs are understood by AI agents across every commerce surface.

Early adopters: L'Oréal, Unilever, Mars, Beiersdorf, and Reckitt.

When five of the world's largest CPG brands move this fast on the same protocol, the rest of the market is already behind.

The Problem: AI Agents Don't Read Your Product Pages

Traditional e-commerce was built for humans browsing product pages.

Agentic commerce removes the page entirely.

When an AI shopping agent executes a purchase on behalf of a user (ChatGPT, Amazon Rufus, Google Gemini, Walmart Sparky), it doesn't load your product detail page. It reasons over product data pulled from the open web, synthesizes information from multiple sources, and makes a recommendation.

If your product data is incomplete, inconsistent, or buried in unstructured content, the agent moves on.

McKinsey projects that by 2030, up to $1 trillion in U.S. B2C retail revenue could be orchestrated through agentic commerce. Globally, the opportunity reaches $3 trillion to $5 trillion.

Morgan Stanley estimates that nearly half of online shoppers will use AI shopping agents by 2030, accounting for approximately 25% of their spending.

The shift is already happening. A 2026 study found that 73% of consumers are using AI in their shopping journey: 45% for product ideas, 37% for summarizing reviews, 32% for price comparison. While only 13% report completing a purchase after an AI referral, 70% are at least somewhat comfortable with an AI agent making purchases on their behalf.

Most brands aren't ready. Their product catalogs were structured for search engines and human browsers, not machine reasoning.

The Infrastructure Gap: Agent Protocols Solve the Wrong Problem

OpenAI's ACP, Google's UCP, AP2, and MCP are agent-side protocols. They solve how agents connect to platforms, execute payments, and access context.

They don't solve the merchant-side problem: how do brands ensure the product intelligence agents consume is accurate, compliant, and strategically governed? ACP and UCP connect agents to product data. They don't control what that data says or where it comes from. That's the gap AMP fills.

What AMP Actually Does

The Agentic Merchant Protocol is merchant-side infrastructure for agentic commerce.

It sits above ACP, UCP, and platform-specific integrations. Brands feed product intelligence into AMP once (catalog data, brand guidelines, compliance rules, regulatory requirements, competitive positioning, persona-level targeting) and AMP handles distribution everywhere.

Four core capabilities:

Canonical product intelligence: Brands create a machine-native source of truth for their catalog (brand guidelines, regulatory guardrails, target personas, competitive context) that all AI agents reference.

Open web distribution: AMP syndicates structured product data to the third-party sources AI agents actually cite: Reddit, Quora, YouTube, Wikipedia, affiliate comparison sites. If an agent pulls data from these surfaces, AMP ensures it is accurate and brand-approved.

Agent-agnostic integration: One protocol covers ChatGPT (ACP), Google Gemini (UCP), Amazon Rufus, Walmart Sparky, Perplexity, and any future agent platform. No lock-in.

Enterprise brand governance: RegGuard Compliance audits all generated content against brand guidelines and regulatory standards before distribution, critical for CPG, pharmaceuticals, and financial services where inaccurate agent-generated claims create legal exposure. Citation tracking shows which sources agents are pulling, so brands can fix misinformation before it scales.

Why L'Oréal, Unilever, Mars, and Beiersdorf Moved First

Max Sinclair, CEO of Azoma:

AMP breaks the foundations of traditional ecommerce. For decades, marketplaces like Amazon and Walmart acted as gatekeepers by controlling product detail pages, rankings, and distribution. In an agentic world, those fixed pages no longer exist.

The traditional model: brands optimized for fixed destinations (pages, ads, search rankings). The agentic model: brands must optimize for dynamic reasoning across distributed sources. When an agent constructs a recommendation, it synthesizes data from dozens of surfaces and weights them by authority and freshness. If the brand's product intelligence isn't structured and distributed across those sources, the agent skips the brand or represents it inaccurately.

The fact that businesses like L'Oréal, Unilever, Mars and Beiersdorf have moved so quickly to adopt AMP tells you everything about the urgency they feel. These are companies that have spent decades building brand equity. They're not about to hand control of how their products are represented to an AI black box.

What Early Adopters Are Seeing

Azoma has been running agentic commerce optimization for CPG brands for three years, before the term existed.

Ruroc (Europe's leading ski helmet brand): Site traffic from ChatGPT increased 14x. Positioned as the number one recommended ski helmet brand in target geographies. AMP ensured agents found structured, extractable, brand-approved product data rather than scraping affiliate blogs.

Matcha brand case (Azoma client): +532% year-over-year revenue growth. AMP syndicated product intelligence to the exact sources AI agents cite when comparing energy alternatives. When the agent reasoned, it encountered structured data built for extraction, not generic marketing copy.

Machine Relations at the Product Level

Brand visibility is downstream of earned media placements in trusted publications that AI engines cite. That's Machine Relations at the brand level, the discipline Jaxon Parrott coined in 2024.

Product visibility follows the same logic: it's downstream of structured, authoritative product intelligence distributed across the sources AI agents trust. The publications haven't changed. AI engines read the same sources (Reddit, Quora, Wikipedia, trusted review sites) that shaped human opinion. What changed is the reader.

Brands that understand this early will define their categories. Brands that wait will find their products represented by whoever got there first.

Three Actions That Matter Now

1. Audit your product data for machine readability. Schema markup correct? Product attributes complete (use cases, certifications, materials)? Content extractable, not buried in JavaScript or image-only tables? If an AI agent tried to recommend your product today based on what it can scrape, would it have enough structured data to do so accurately?

2. Map where AI agents are citing you (or not). Use a citation tracking tool to see which AI platforms mention your brand, which queries trigger recommendations, which sources the agents cite, and where competitors appear while you don't. You can't improve visibility until you know where you're invisible.

3. Decide: build in-house or adopt a protocol. Integrating directly with ACP, UCP, and every future agent platform while manually distributing product data to citation sources is viable only for platforms large enough to set their own standards (Amazon, Walmart, Shopify). For everyone else, adopting AMP or an equivalent merchant-side protocol means one integration and universal distribution instead of fragmented in-house builds that perpetually lag the market.

The Architecture Is the Same

PR practitioners are concluding that AI citations are the new success metric.

GEO researchers are proving that earned media is what AI engines actually cite.

Product teams are discovering that AI agents synthesize distributed data rather than browsing brand-owned pages.

All three observations are fragments of the same shift: machines now mediate discovery, and the brands that govern how machines represent them will define their categories.

Machine Relations is the architecture that connects these fragments. At the brand level, it's earned media in trusted publications. At the product level, it's structured product intelligence distributed across the sources agents trust. Both solve the same question: when someone asks an AI system about your category, do you show up as the authoritative answer?

L'Oréal, Unilever, Mars, Beiersdorf, and Reckitt bet yes and moved first.

The question now is how long the rest of the market waits.

Want to see how your brand shows up in AI answers today?

Run a free AI Visibility Audit at app.authoritytech.io/visibility-audit and see where you're cited, where competitors appear, and which AI platforms are recommending you (or not).