Feedonomics Launches Agentic Catalog Exports: How Products Enter the AI Shopping Pipeline

Originally published on The Searchless Journal

On April 27, Feedonomics, a subsidiary of Commerce (Nasdaq: CMRC), launched Agentic Catalog Exports (ACE), a product data syndication platform purpose-built for AI shopping agents. Dell is the flagship customer. The initial integration targets seven surfaces: OpenAI/ChatGPT, Google Gemini, Microsoft Copilot, PayPal, Stripe, Perplexity, and Amazon.

ACE is not another channel feed. It is the supply-side infrastructure layer that makes agentic commerce operational. Traditional feeds push product data to surfaces where humans browse and click. ACE pushes structured, enriched product data to surfaces where AI agents query, compare, and recommend on behalf of human buyers. The distinction sounds subtle. The business implications are not.

What ACE Actually Does

Feedonomics has spent years as a feed management platform, helping merchants normalize and syndicate product data to Google Shopping, Amazon, Meta, and dozens of comparison shopping engines. ACE extends that core capability into a new domain: AI agent consumption.

The platform takes raw merchant catalog data and transforms it through three layers:

Structural enrichment. Raw catalog data from ERPs and PIMs is notoriously incomplete. Missing GTINs, inconsistent categorization, blank attribute fields. ACE fills gaps and normalizes formats so that AI agents receive complete, queryable records rather than half-empty product pages.

Semantic mapping. A product title like "XPS 15 9530 Intel i7-13700H 32GB 1TB RTX 4060" means nothing to a consumer asking ChatGPT "what's a good laptop for video editing under $2,000?" ACE maps technical specifications to use cases, intent signals, and comparative benchmarks. The agent receives not just specs but context: this machine handles 4K timelines in DaVinci Resolve, compares favorably to the MacBook Pro 14" on price-to-performance, and is overkill for web browsing.

Protocol适配. Each AI surface has different ingestion requirements, data formats, and API specifications. Google's Merchant Center schema differs from what OpenAI expects in a structured product graph. PayPal's shopping recommendation engine wants different attributes than Perplexity's product comparison module. ACE handles the translation layer so merchants maintain one enriched catalog and syndicate everywhere.

The result: when a consumer asks an AI agent for a product recommendation, the agent queries a structured database where your products either exist with rich, accurate attributes or they don't exist at all. ACE determines whether you're in that database.

The AI Shopping Surfaces: Who's Building What

ACE targets seven platforms. Each has a different strategic position and data appetite:

Platform	User Base (Est.)	Shopping Capability	Data Ingestion Method	Strategic Priority for Merchants
OpenAI/ChatGPT	900M WAUs	Product research, recommendations, affiliate links	Structured product graph + crawling	Critical. Largest AI surface by reach.
Google Gemini	400M+ MAUs	Shopping in Search, Gemini assistant, Lens	Merchant Center + Gemini API	Critical. Ties to existing Google Shopping infra.
Microsoft Copilot	100M+ MAUs	Shopping in Bing, Edge, Windows	Product feeds + Bing Shopping	High. Growing distribution via Windows default.
PayPal	400M+ active accounts	AI-driven recommendations, deal discovery	Structured feed partnerships	Medium-High. Massive buyer intent data.
Stripe	Millions of merchants	Checkout-level AI recommendations	Product API integration	Medium. Early stage but high purchase intent.
Perplexity	100M+ MAUs	AI-native product search and comparison	Structured product data + crawling	High for tech-savvy demographics.
Amazon	300M+ active customers	Rufus AI assistant, product Q&A	Existing marketplace data + feed enrichment	Critical for categories already on Amazon.

Seven surfaces, seven different data requirements, seven different consumer contexts. The merchant who tries to build custom integrations for each one faces an engineering nightmare that multiplies with every new surface. ACE abstracts this complexity into a single pipe.

Sharon Gee, SVP of Product for AI at Commerce, framed it plainly: "Merchants need a reliable way to participate. With ACE, we're making it easier for enterprises to prepare their product data without having to build and maintain complex, one-off integrations for every destination."

Traditional Feeds vs. Agent-Ready Feeds

The gap between a traditional product feed and an agent-ready feed is not incremental. It is categorical.

Dimension	Traditional Feed (Google Shopping, Meta)	Agent-Ready Feed (ACE)
Primary consumer	Human eyes browsing a grid	AI agent parsing structured data
Data model	Title, description, price, image, GTIN	Structured attributes, use cases, comparisons, specs
Content optimization	Keyword-stuffed titles, marketing copy	Clean specs, intent mapping, compatibility matrices
Attribute depth	20-50 fields per product	100+ fields including contextual metadata
Update frequency	Daily or hourly batch	Near real-time or on-demand query
Discovery mechanism	Keyword match, bid ranking	Semantic similarity, intent matching
Conversion path	Click → land → browse → add to cart	Agent recommends → one-click buy (UCP)
Competitive dynamic	Highest bid + best listing wins	Richest, most accurate data wins

The shift from keyword-match to semantic-match is the defining change. In Google Shopping, a merchant can win placement by bidding aggressively on high-intent keywords. In an AI agent context, the agent evaluates product fitness against the user's specific requirements. You cannot bid your way to a recommendation. You can only be the best-structured answer.

This is why Feedonomics positions ACE not as a channel add-on but as a new infrastructure category. The optimization rules are fundamentally different.

Dell's Adoption: A Case Study in Enterprise Readiness

Dell is ACE's launch customer, with approximately 7,000 SKUs prepared for agent-driven discovery. The product catalog spans laptops, desktops, servers, monitors, and accessories, a deliberately complex test case.

Paul Mansour, Dell's Global Marketing Director, explained the logic: "As AI agents become a more common starting point for product discovery, the quality and structure of product data matter more than ever. Feedonomics helped us optimize and structure our catalog so Dell products are not only more discoverable, but also more accurately and completely represented within ChatGPT."

Three things make Dell's adoption strategically interesting:

1. Hardware requires rich specification data to recommend well. A t-shirt or a candle needs a title, price, image, and maybe a size. A server recommendation requires processor architecture, RAM configuration, storage tier, rack units, power consumption, thermal output, supported workloads, and compatibility with existing infrastructure. Dell's product portfolio sits at the extreme end of specification complexity. If ACE can handle 7,000 Dell SKUs with accurate technical enrichment, it can handle most merchant catalogs.

2. Dell's direct-to-consumer model depends on AI discovery. Dell sells direct. It does not have the Amazon marketplace flywheel pushing its products to the top of search results. As consumers shift from "googling and clicking" to "asking an AI agent," Dell's discoverability hinges entirely on whether its products appear in agent recommendations. ACE is not a nice-to-have for Dell. It is a survival strategy for the direct sales model.

3. The Dell signal will accelerate enterprise adoption. When a Fortune 500 company with a $50B+ annual revenue puts its name on a press release for a new product category, it shortens the evaluation cycle for every other enterprise e-commerce VP. Expect HP, Lenovo, and other hardware manufacturers to follow within months. Expect consumer electronics, appliances, and automotive parts categories to follow within a year.

Dell's move is not experimental. Dell specifically called out ChatGPT representation in its quote, indicating live production usage, not a pilot. The company has committed engineering resources to structuring 7,000 products for agent consumption. This is operational infrastructure.

The Competitive Landscape: Feed Wars Are Coming

Feedonomics is early with a dedicated agentic feed product, but the feed management space is crowded and well-funded. The question is not whether competitors will respond, but how fast.

Google Merchant Center already has the largest product data corpus on the internet. Google's incentive is to make product data available to Gemini and Google AI surfaces natively. Google does not need Feedonomics as an intermediary for its own surfaces. Its vulnerability is that Google's schema was built for Shopping ads, not for agent consumption. The enrichment gap is real.

Shopify controls product data for millions of SMBs. The company has been exploring agentic storefront integrations, including partnerships with ChatGPT for product discovery. Shopify's advantage is distribution: any feature shipped to its merchant base reaches scale instantly. Its disadvantage is that Shopify merchants skew small, with product data quality to match. Agent-ready data requires enrichment that many Shopify merchants cannot produce themselves.

ChannelAdvisor (now Rithum) and GoDataFeed are traditional feed management platforms adapting to the AI era. They have existing merchant relationships and channel expertise but have not yet shipped a dedicated agentic product. Expect announcements within Q2-Q3 2026.

Amazon is the sleeping giant. Amazon already has the richest product data corpus in the world, and its Rufus AI assistant is live. Amazon's incentive is to keep product data inside its walled garden, not to syndicate it to OpenAI and Google. But Amazon also sells third-party products via its marketplace, and those sellers want their products recommended by every AI agent, not just Rufus. Tension between Amazon's platform control and sellers' multi-surface ambitions will define this category.

The likely outcome: Feedonomics has a 6-12 month first-mover window before Google, Shopify, and others close the gap. Enterprise merchants with complex catalogs (hardware, B2B, industrial) will gravitate toward Feedonomics because enrichment depth matters more for technical products. SMB merchants will eventually get agent-ready feeds through Shopify or Google natively. The middle market, merchants too large for Shopify's basic tools but without Dell-level engineering budgets, is the contested territory.

The UCP Connection: Supply Meets Demand

ACE does not exist in a vacuum. The Universal Commerce Protocol (UCP), which Google began operationalizing in March 2026 with Cart, Catalog, and Identity capabilities, is the demand-side standard for how AI agents initiate and execute purchases. ACE is the supply-side complement.

Together, they form the agentic commerce stack:

• ACE (supply side): Ensures product catalogs are structured, enriched, and syndicated to AI surfaces. Products become discoverable and evaluable by agents.
• UCP (demand side): Standardizes how agents identify products, check inventory and pricing, and execute transactions across platforms. Agents can actually buy what they find.

Without ACE, UCP-enabled agents find products but work with incomplete or poorly structured data, leading to inaccurate recommendations. Without UCP, ACE-syndicated products are discoverable but lack a standardized purchase protocol, forcing agents into brittle, platform-specific checkout flows.

Merchants need both. The good news: they are converging. Feedonomics has explicitly designed ACE for "agentic protocol-driven destinations." The infrastructure is being built to interoperate.

Who Wins, Who Loses

Agentic commerce will create winners and losers at every layer of the e-commerce stack.

Winners:

• Enterprises with clean, structured product data. Companies that have invested in PIM systems, attribute management, and data governance will feed cleanly into ACE and similar platforms. Their products will be recommended accurately from day one.
• Technical product categories. Hardware, electronics, B2B equipment, automotive parts, categories where purchase decisions depend on specifications rather than aesthetics. Agent-ready feeds favor rich attribute data.
• Direct-to-consumer brands. DTC brands that lack marketplace distribution can use agent discovery to bypass Amazon's flywheel. If ChatGPT recommends your product directly, you don't need Amazon placement.
• Feedonomics/Commerce. First-mover advantage with a purpose-built product and a Fortune 500 customer on the record. The CMRC stock will trade on agentic commerce narrative.

Losers:

• Merchants with dirty catalog data. If your product titles are inconsistent, your attributes are missing, and your categorization is a mess, no amount of feed syndication will fix the underlying problem. ACE enriches; it does not perform miracles.
• Amazon-dependent sellers. If your entire distribution runs through Amazon's marketplace, you have optimized for Amazon's search algorithm, not for agent-ready structured data. The skills and data models are different.
• SEO-only merchants. Traditional organic search optimization targets Google's web crawler. Agent optimization targets a different set of systems with different requirements. The overlap is smaller than most SEO professionals want to admit.
• Traditional comparison shopping engines. Shopzilla, PriceGrabber, and their descendants optimized for human browsing grids. Agent-native surfaces render them obsolete.

A Contrarian Take: Agentic Commerce Is Overhyped in the Short Term

The McKinsey $1 trillion estimate for agentic commerce by 2030 gets cited constantly. It deserves scrutiny.

Capgemini's survey finding that 38% of consumers "trust AI agents for routine purchases" sounds impressive until you parse what "routine purchases" means: reordering toothpaste, repurchasing a specific brand of coffee, auto-replenishing cleaning supplies. These are low-complexity, low-risk transactions where the agent is executing a known preference, not making a discovery decision.

The hard part of commerce, and where the $1 trillion actually resides, is consideration purchases: products the consumer has never bought before, categories where they need to compare options, price points where mistakes are costly. AI agents are nowhere near reliable enough for most consumers to trust with a $1,500 laptop recommendation without verifying the agent's suggestion independently. Which means the consumer is still doing the research themselves. Which means the agent is an input, not a decision-maker.

The honest timeline:

• 2026-2027: Agents handle replenishment and simple reorders well. Discovery recommendations are advisory, not authoritative. Product data preparation is necessary but early volume is modest.
• 2028-2029: Agents become competent at consideration purchases in structured categories (electronics, appliances, travel). Consumer trust grows. Feed quality directly correlates with recommendation volume.
• 2030+: Agents handle most purchase decisions in most categories. Agentic commerce is a primary channel, not a specialty channel. The $1 trillion estimate becomes plausible.

The contrarian position: merchants should invest in agent-ready data now not because the revenue is imminent, but because the data quality improvements pay for themselves across every channel. Better product attributes improve your Google Shopping performance, your marketplace listings, your on-site search, and your customer service accuracy today. Agent readiness is a bonus, not the sole justification.

Feedonomics knows this. ACE is positioned as an extension of existing feed optimization, not a standalone product. The pitch to merchants is not "spend money on a speculative future channel." It is "improve your data and get agent readiness for free." Smart framing.

The Merchant Playbook: How to Prepare

Whether or not you adopt ACE specifically, the data preparation work is the same. Here is a framework:

Phase 1: Data Audit (Weeks 1-4)

Inventory your catalog's attribute completeness. For each product, measure:

• Percentage of required fields populated (GTIN, brand, MPN, category)
• Attribute depth (how many specification fields exist per product)
• Data consistency (same attribute, different formats across products)
• Image quality and completeness (multiple angles, lifestyle, contextual)

Benchmark against your three strongest competitors. If their data is richer than yours, you lose in agent recommendations regardless of product quality.

Phase 2: Enrichment (Weeks 5-12)

Fill the gaps identified in the audit. Prioritize by:

• Revenue per SKU (enrich high-value products first)
• Search volume for the product category (high-demand categories yield faster returns)
• Agent surface relevance (categories where AI recommendations are already live, like electronics and software, should go first)

For each product, add:

• Technical specifications in structured format (not buried in description text)
• Use case tags (what jobs this product solves)
• Compatibility data (what it works with)
• Comparative positioning (where it sits relative to alternatives)

Phase 3: Syndication (Weeks 13-16)

Choose your distribution path:

• Enterprise merchants (10,000+ SKUs): Evaluate ACE and similar managed platforms. The engineering savings justify the cost.
• Mid-market merchants (1,000-10,000 SKUs): Test with Google Merchant Center's emerging AI features first. Add dedicated agent feeds as surfaces mature.
• SMB merchants (<1,000 SKUs): Focus on Google Shopping and Shopify's native tools. Agent-specific optimization is premature for most small catalogs.

Phase 4: Measurement (Ongoing)

Track agent-driven traffic and conversions separately from traditional channels. Key metrics:

• Agent surface impressions (how often your products appear in AI recommendations)
• Agent click-through rates (how often consumers act on AI suggestions)
• Agent-attributed revenue (direct conversion from agent recommendations)
• Competitive visibility (how often your products appear vs. competitors in identical queries)

Most analytics platforms do not distinguish agent traffic from organic yet. Build custom UTM parameters and landing page tracking now so you have baseline data when the volume scales.

The GEO Dimension: Product-Level AI Visibility

For brands investing in Generative Engine Optimization, ACE introduces a new variable. Traditional GEO focuses on content citation: ensuring your brand's expertise, thought leadership, and informational content are surfaced by AI engines. ACE operates at the product level: ensuring individual SKUs are discoverable and accurately represented.

Both matter. Content citation builds brand trust and awareness. Product discovery drives purchase consideration. A brand that ranks first in ChatGPT's answer to "what's the best project management software?" but has no product data in ChatGPT's shopping database loses the conversion to a competitor with worse content but better-structured product feeds.

The winning strategy operates on both dimensions simultaneously:

1. Content GEO: Publish authoritative, well-structured content that AI engines cite when answering category-level questions. This builds the brand as a trusted answer.
2. Product GEO: Syndicate enriched, agent-ready product data through ACE and similar platforms. This ensures that when the agent transitions from "what should I buy?" to "which specific product?" your SKUs are in the consideration set.

Most brands are investing heavily in #1 and ignoring #2 entirely. ACE makes #2 accessible without bespoke engineering.

What Comes Next

ACE launches as an enterprise service with manual onboarding. Feedonomics has signaled that self-service capabilities are coming, along with mid-market accessibility. The trajectory:

1. Current state: Enterprise-only, managed service, seven surfaces, Dell as proof point.
2. H2 2026: Self-service tier for mid-market. Expanded surface coverage as new AI shopping platforms emerge.
3. 2027: Commodity layer. Every major feed management platform offers agent-ready syndication. Differentiation shifts to enrichment quality and surface-specific optimization.
4. 2028+: Agent-ready feeds are table stakes. The competitive battleground moves to real-time pricing, inventory, and personalized recommendation signals.

The window where agent-ready feeds are a competitive advantage is measured in months, not years. The window where agent-ready feeds are a competitive necessity is measured in years, not decades. The cost of inaction compounds: every month your products are absent from agent databases, those databases build patterns around your competitors' products instead.

Commerce's stock (CMRC) will trade on the agentic commerce narrative for the rest of 2026. The question for investors is whether ACE represents durable competitive advantage or a temporary first-mover position that Google and Shopify erode within 18 months. The answer depends on whether Feedonomics' enrichment layer, the transformation of raw catalog data into agent-ready structured data, is genuinely hard to replicate or just hard to replicate first.

Based on Dell's adoption and the explicit technical enrichment requirements for complex product catalogs, Feedonomics has a defensible position in enterprise, where data complexity creates a moat. In SMB, where product catalogs are simpler and Google/Shopify have direct merchant relationships, the moat is thinner.

---

Check If Your Products Are Discoverable by AI Shopping Agents

See how your catalog performs across ChatGPT, Gemini, Copilot, and other AI shopping surfaces. Our audit reveals your product data readiness, attribute gaps, and specific optimization opportunities for agentic commerce.

Start Your Free AI Visibility Audit

---

Sources

• GlobeNewswire: Feedonomics Agentic Catalog Exports press release (April 27, 2026)
• Commerce (Nasdaq: CMRC) official announcement and investor materials
• ChannelX: Feedonomics ACE coverage and analysis (April 27, 2026)
• SiliconANGLE: Google Cloud Next "invisible shelf" concept (April 27, 2026)
• Capgemini: Consumer trust in AI agents for purchases (April 2026)
• McKinsey/ICSC: Agentic commerce revenue projections (April 24, 2026)
• Gladly: Agentic commerce protocols (MCP, ACP, UCP) explained (April 23, 2026)
• CBS News: AI shopping agents consumer coverage (April 2026)
• Search Engine Land: Top e-commerce trends 2026 (April 2026)
• Searchless Journal: Universal Commerce Protocol coverage (April 25, 2026)

---

FAQ

What is Agentic Catalog Exports (ACE)?

ACE is a Feedonomics platform that transforms e-commerce product catalogs into structured, enriched data formats optimized for AI agent consumption. It syndicates agent-ready product data to seven AI shopping surfaces: ChatGPT, Google Gemini, Microsoft Copilot, PayPal, Stripe, Perplexity, and Amazon.

How is agent-ready product data different from traditional product feeds?

Agent-ready data prioritizes structured attributes (specs, use cases, compatibility, comparative data) over marketing copy. Traditional feeds are designed for humans browsing a grid; agent-ready feeds are designed for AI systems parsing structured data to make recommendations. The attribute depth is typically 2-3x greater than a standard Google Shopping feed.

Why is Dell's adoption significant?

Dell is syndicating approximately 7,000 SKUs through ACE, spanning laptops, desktops, servers, monitors, and accessories. Hardware is one of the hardest categories for agent-ready enrichment due to the depth of technical specifications required. Dell's live production usage validates that ACE handles enterprise-scale complexity. Dell's direct-to-consumer model also makes agent discovery strategically critical.

What is the "invisible shelf"?

The invisible shelf refers to structured product databases that AI agents query when making recommendations. Unlike traditional e-commerce where products sit on visible pages humans browse, the invisible shelf is machine-readable inventory that consumers never directly see. Your products either exist on it with accurate data or they are invisible to AI-driven commerce.

How does ACE relate to UCP (Universal Commerce Protocol)?

ACE is the supply-side complement to UCP. ACE ensures products are structured and discoverable in agent databases. UCP standardizes how agents execute purchases across platforms. Merchants need both: ACE so agents can find and recommend their products, and UCP so agents can complete the transaction.

When should merchants invest in agent-ready product data?

Now, but for the right reasons. Data quality improvements (attribute completeness, structured specs, use case mapping) benefit every channel, not just AI surfaces. The ROI from better Google Shopping performance, improved marketplace listings, and more accurate on-site search pays for the enrichment work. Agent readiness is the bonus that compounds over time as AI shopping volume scales.

---

Prepare Your Products for AI Shopping

Agentic commerce is the future of product discovery. Learn how to make your catalog AI-ready and capture the next wave of shopping volume.

Explore AI Visibility for Ecommerce