Research

Agent-Readable Product Feeds, May 2026

How brands should structure product feeds for AI agent consumption in 2026. Google Merchant Center, Schema.org Product, ChatGPT Shopping feeds, Perplexity product cards, and emerging agent-feed standards compared.

By Ramanath, CTO & Co-Founder at Presenc AI · Last updated: May 2026

Why Product Feeds Matter for Agent Marketing

When an AI agent constructs a candidate set for a purchase or comparison query, the candidate set comes from one of four sources: (1) the model's training corpus, (2) live search retrieval, (3) a structured product feed the brand has submitted to the platform, or (4) an MCP server the brand has published. Product feeds are the cheapest, most predictable, and most under-invested of the four. This page consolidates the agent-readable feed standards in May 2026.

Product Feed Surface Comparison

SurfaceFormatConsumer OfStatus
Google Merchant CenterXML / Google product feed specGoogle Search, Shopping, Gemini ShoppingProduction, ~20-year history
Schema.org Product (JSON-LD)Embedded on product pagesAll major LLM crawlers + search enginesProduction, open standard
ChatGPT Shopping feedOpenAI-defined JSONChatGPT product recommendationsProduction (launched late 2024)
Perplexity product cardsSchema.org Product + Perplexity-specific extensionsPerplexity ShoppingProduction
Microsoft Merchant CenterXML compatible with Google feedBing, Bing Copilot ShoppingProduction
MCP "shopping" serversMCP tool-call interface (Anthropic standard)Claude Agent, custom agent stacksEmerging
Agent Pay product manifestAP2-compatible product declarationAP2-integrated agents (Google Gemini Shopping, others)Production

Critical Schema.org Product Properties for Agent Extraction

PropertyImportanceNotes
nameRequiredExact product name as displayed
brandCriticalBrand name; supports cross-product brand recall
offers (Offer)CriticalpriceCurrency, price, availability (InStock/OutOfStock), priceValidUntil
sku / mpn / gtinHighUniversal product identifiers; help dedupe across feeds
descriptionHigh3-5 sentence concise description; fact-dense
aggregateRatingHighratingValue, reviewCount; drives agent confidence
image (multiple)MediumMultiple URLs at multiple resolutions
review (multiple Review)MediumIndividual reviews with author + reviewRating
categoryHighGoogle product taxonomy preferred; aids agent routing
weight, height, depth, widthMediumQuantitativeValue; matters for shipping-related queries
shippingDetailsMediumOfferShippingDetails with delivery time + cost
hasMerchantReturnPolicyMediumReturn-policy details; agent purchase decisions weight this

Six Things the Feed Landscape Tells You

  1. Schema.org Product on owned pages is the single most leveraged investment. One implementation produces extraction signal across every major LLM crawler, every search engine, and every agent stack. Brands with strong Schema.org coverage are extractable everywhere; brands without it are dependent on each platform's individual feed submission, which is fragmented and labor-intensive.
  2. Google Merchant Center remains the highest-volume feed surface. ~20-year ecosystem maturity, billions of products in catalog, and now feeds directly into Gemini Shopping. For commerce brands, Merchant Center completeness is table stakes.
  3. ChatGPT Shopping feed is the recent high-leverage entrant. Launched late 2024, the OpenAI-defined JSON feed is the only path to ChatGPT product recommendations. Submission is direct and underbuilt, which means competitive moats are still available to early movers.
  4. Bing Copilot inherits Merchant Center automatically. Brands submitting to Google Merchant Center can submit the same feed (with minor tweaks) to Microsoft Merchant Center for Bing Copilot Shopping. Low-incremental-cost coverage extension.
  5. MCP shopping servers are the emerging agent-side path. A branded MCP server that exposes product inventory, pricing, and availability gives agentic AI (Claude Agent, custom stacks) direct programmatic access. Currently very underbuilt; high-leverage for any brand with substantial inventory.
  6. Agent Pay product manifests are required for AP2 agent commerce. AP2-mediated agent purchases require AP2-compatible product declarations. Brands selling through AP2-integrated agents must publish these manifests to be eligible for agent-mediated purchase.

What This Means for Brand AI Visibility

Product feed completeness is now a structural input to commerce-related AI visibility. Brands selling through any AI-mediated shopping surface need: (1) comprehensive Schema.org Product on owned pages, (2) Google Merchant Center submission, (3) ChatGPT Shopping feed submission, (4) Microsoft Merchant Center (incremental from Google feed), (5) optionally an MCP shopping server and AP2 product manifest for agentic surfaces. The combined investment is moderate but the visibility lift compounds across surfaces. Brands that skip feed work cede agent-mediated commerce candidate-set inclusion to competitors who invested.

Methodology

Feed format details collected May 15, 2026 from vendor documentation (Google Merchant Center, Schema.org Product spec, OpenAI ChatGPT Shopping documentation, Microsoft Merchant Center). MCP and AP2 specifications from the Linux Foundation Agentic AI Foundation (AAIF) publications. Refreshed quarterly as feed surfaces continue to launch and evolve.

How Presenc AI Helps

Presenc AI tracks brand presence inside AI-mediated shopping surfaces (ChatGPT product recommendations, Perplexity Shopping, Gemini Shopping, Bing Copilot Shopping). When a brand's product fails to appear in a relevant agent candidate set, our instrumentation traces back to which feed source produced the gap.

Frequently Asked Questions

Start with Schema.org Product JSON-LD on every product page (one implementation, broad extraction signal). Then submit to Google Merchant Center, Microsoft Merchant Center, and ChatGPT Shopping. For agentic surfaces, add an MCP shopping server and AP2 product manifest. The combined surface coverage takes 1-3 sprints to build but compounds visibility across every AI-mediated shopping channel.
Yes, with submission requirements. The OpenAI-defined JSON feed launched late 2024 and is the only path to ChatGPT product recommendations. Submission is direct and currently underbuilt relative to its potential reach; early submitters get disproportionate visibility in product-recommendation queries.
High. AI agents constructing candidate sets weight aggregateRating heavily as a confidence signal. Products without rating data (ratingValue + reviewCount) often surface lower than competitors with even modest rating completeness. For brands not generating sufficient first-party reviews, integrating with G2, Capterra, Trustpilot, or App Store review data and surfacing aggregateRating in your feed is high-leverage.
Optional but increasingly recommended for inventory-heavy brands. Traditional feeds (Google Merchant Center, Schema.org, ChatGPT Shopping) cover the bulk of consumer-facing agent surfaces. MCP shopping servers cover the agentic AI surface (Claude Agent, custom enterprise agents) which is smaller today but growing fast. Brands with API-accessible inventory should publish an MCP server in 2026 to claim the surface before it's fully competed.

Track Your AI Visibility

See how your brand appears across ChatGPT, Claude, Perplexity, and other AI platforms. Start monitoring today.