Product Feed Optimization for AI Shopping Agents: How to Get Your Catalog Picked in Google AI Mode, ChatGPT and Perplexity?

Black Friday falls on November 27 this year and Cyber Monday on November 30. If you’re reading this the week it was published, you have about 18 days.

That’s more than enough time to fix your product feed. It is not enough time to onboard a new agentic checkout program, so this guide splits the work into two lists: what to do before Black Friday, and what to start in the first quarter of 2027.

Here’s why the feed comes first. Google’s own holiday guidance for retailers says holiday success “ultimately relies on an accurate product feed,” and reports that merchants adopting its core Merchant Center feed best practices see, on average, a 5% increase in conversions the following month (Google). Adobe expects traffic from AI platforms to US retail sites to grow 130% year over year this season (Search Engine Watch). The agents sending that traffic read your feed before they ever read your page.

What Is Product Feed Optimization for AI?

Product feed optimization for AI is the work of making your catalog data complete, accurate, and current enough that AI shopping agents can match your products to detailed requests, compare them fairly against alternatives, and either send the shopper to buy or complete the purchase.

What’s different from classic Shopping feed work is the query. A shopper used to type “men’s running shoes.” Now they ask an assistant for “waterproof trail shoes for wide feet under $120 that can arrive before Saturday.” Every clause in that request maps to an attribute: waterproof, trail, wide, price, delivery date. Missing attributes mean missing matches.

For the bigger picture of how agents discover, compare, and buy, read our guide to agentic commerce explained.

Where Each AI Platform Gets Product Data?

Sources: Google Merchant Center Help on UCP, Google on agentic checkout, OpenAI on product discovery, PayPal and Perplexity, Copilot Checkout.

Two details are easy to miss. First, OpenAI’s feed reference says standard uploads currently target the United States, with Canada and Mexico available through additional market setup (OpenAI). Second, Google’s shopping data is enormous: more than 50 billion product listings in the Shopping Graph, with 2 billion updated every hour (Google). Your products compete inside that pool on data quality.

The Agent Critical Attributes

This table maps what an agent needs to the field names on Google Merchant Center and in OpenAI’s ACP product feed. Field names are shown exactly as the platforms define them.

ACP checkout also requires public seller_privacy_policy and seller_tos URLs, and OpenAI notes that setting the eligibility flag doesn’t complete checkout onboarding on its own (OpenAI). On Google, only listings that use the native_commerce attribute will show the “Buy” button for UCP powered checkout (Google Merchant Center Help).

Merchant Center Conversational Attributes, Explained

In 2026 Google added a set of optional conversational attributes designed to help AI systems understand product nuances on surfaces like AI Mode. Google recommends submitting them through a supplemental data source, says they won’t affect the approval status of existing products, and notes you don’t need to repeat details already in your description, product highlights, or product details (Google Merchant Center Help).

Do these work? Google says that in testing with lululemon, conversational attributes submitted by the brand were incorporated 50% of the time in relevant product recommendations in AI Mode (Google). That’s one brand and Google’s own figure, so treat it as directional. It’s still the clearest evidence available that the attributes get used.

Step by Step: Make Your Feed Agent Ready

1. Fix mismatches first. Price, availability, and shipping must match between your feed and your landing page. Agents compare options in near real time, and a mismatch gives them a reason to choose someone else. It can also get products disapproved.
2. Rewrite titles for matching, not persuasion. Use brand, product type, key attributes, then variant. “Northline Waterproof Trail Running Shoe, Men’s Wide, Black, Size 10” matches far more requests than “Best Shoes for Your Next Adventure!” Keep promotional text out of titles; OpenAI’s feed reference suggests aiming for at most 150 characters.
3. Write descriptions like spec sheets with context. Materials, dimensions, fit, compatibility, use cases, care, and what’s in the box, in plain text. Assume the agent will quote you.
4. Complete your identifiers. Add GTINs wherever they’re assigned, plus brand and MPN. OpenAI specifies that a GTIN must be 8, 12, 13, or 14 digits with a valid check digit and no spaces or dashes.
5. Model variants properly. Send one row per variant with a stable ID, and group variants under a parent. OpenAI’s reference is explicit: never include price in an offer ID, and keep IDs stable when price, stock, title, or images change.
6. Publish shipping and returns everywhere agents look. Use Merchant Center shipping and return settings, show the terms as plain text on product pages, and consider Organization level structured data. Google supports a store wide return policy through hasMerchantReturnPolicy and a store wide shipping policy through hasShippingService (Google; Google). Settings in Search Console and Merchant Center take precedence over on page markup.
7. Add conversational attributes for your top sellers. Start with question and answer data for your top 50 products, using the questions customers actually ask, then add related products for accessories and required parts.
8. Get reviews into the data. Join Merchant Center’s Product Ratings program, and in an ACP feed send review count and star rating that describe the same review population.
9. Speed up updates for the season. Increase scheduled fetches or use the API so price and stock changes land within hours, not days. OpenAI notes that uploaded sale dates don’t schedule price changes, so update the current price when it changes.
10. Read the AI performance insights report. If your Merchant Center account is in Australia, Canada, India, New Zealand, or the US, this report shows your share of voice in AI Mode and AI Overviews against defined competitors, top terms shoppers use, and popular attributes your data may be missing (Google Merchant Center Help). Fill the gaps it surfaces, starting with the most frequent terms.

Before and After: Product Page Markup

Your product page still matters. Agents and crawlers use it to verify the feed, and Google’s guidelines require structured data to match what’s visible on the page (Google). Structured data isn’t required to appear in AI Mode, according to Google (Google), but complete Product markup supports merchant listing eligibility and makes your facts easier to confirm. For the wider schema picture, see our guide to schema markup for AI Mode.

Before

{

“@context”: “https://schema.org”,

“@type”: “Product”,

“name”: “Trail Runner”,

“offers”: { “@type”: “Offer”, “price”: “119” }

}

What’s wrong: no brand, no identifier, no currency, no availability, no image, no shipping or returns, and no ratings. An agent can confirm almost nothing.

After

{

“@context”: “https://schema.org”,

“@type”: “Product”,

“name”: “Northline Waterproof Trail Running Shoe, Men’s Wide, Black”,

“sku”: “TR_BLK_W_10”,

“gtin13”: “4006381333931”,

“brand”: { “@type”: “Brand”, “name”: “Northline” },

“image”: “https://example.com/img/trailrunner_black.jpg”,

“description”: “Waterproof trail running shoe with a wide toe box and a rubber lug outsole.”,

“aggregateRating”: { “@type”: “AggregateRating”, “ratingValue”: “4.6”, “reviewCount”: “254” },

“offers”: {

“@type”: “Offer”,

“url”: “https://example.com/products/trailrunner?size=10”,

“price”: “119.00”,

“priceCurrency”: “USD”,

“availability”: “https://schema.org/InStock”,

“itemCondition”: “https://schema.org/NewCondition”,

“shippingDetails”: {

“@type”: “OfferShippingDetails”,

“shippingRate”: { “@type”: “MonetaryAmount”, “value”: “0”, “currency”: “USD” },

“shippingDestination”: { “@type”: “DefinedRegion”, “addressCountry”: “US” },

“deliveryTime”: {

“@type”: “ShippingDeliveryTime”,

“handlingTime”: { “@type”: “QuantitativeValue”, “minValue”: 0, “maxValue”: 1, “unitCode”: “DAY” },

“transitTime”: { “@type”: “QuantitativeValue”, “minValue”: 2, “maxValue”: 4, “unitCode”: “DAY” }

}

},

“hasMerchantReturnPolicy”: {

“@type”: “MerchantReturnPolicy”,

“applicableCountry”: “US”,

“returnPolicyCategory”: “https://schema.org/MerchantReturnFiniteReturnWindow”,

“merchantReturnDays”: 30,

“returnMethod”: “https://schema.org/ReturnByMail”,

“returnFees”: “https://schema.org/FreeReturn”

}

}

}

Every value in the markup has to appear on the page too: the price, the rating, the delivery window, and the return terms. If your return and shipping policies are the same across the store, define them once at the Organization level instead of repeating them on every product.

Market Notes: US, UK, and India

• United States: Everything in this guide applies. UCP powered checkout (for select merchants), agentic checkout, ChatGPT feeds, Copilot Checkout, and Perplexity’s Instant Buy all started here.
• United Kingdom: Focus on Merchant Center quality and on your product pages. UCP checkout eligibility currently lists the US, Canada, and Australia, Merchant Center’s AI performance insights don’t currently list the UK, and ChatGPT’s standard feed uploads target the US. Use GBP pricing that follows UK VAT display conventions, and keep feeds clean so you’re ready when programs expand.
• India: AI performance insights are available for English queries, so use them. OpenAI’s current feed markets don’t include India. UPI based agentic payments have been piloted in ChatGPT with Razorpay and NPCI since October 2025 (Razorpay), while NPCI’s broader protocol was put on hold in September 2026 pending regulatory approval (Business Standard). Prioritize Merchant Center accuracy, INR pricing, and delivery promises by location.

Black Friday and Cyber Monday 2026 Preflight Checklist

By November 13

• Clear every Merchant Center disapproval and item warning on your top sellers.
• Confirm price and availability match between feed and landing pages for your top 200 products.
• Load planned Black Friday prices as sale prices and test on a handful of items first.
• Publish holiday shipping cutoffs and extended return windows as plain text on pages and in Merchant Center settings.

By November 20

• Add question and answer data and related products for your top 50 products through a supplemental data source.
• Check that product ratings are flowing into Merchant Center.
• Increase feed update frequency for price and stock, or switch to API updates.
• Review AI performance insights for missing attributes and top terms, if your account is eligible.
• If you’re on Shopify, confirm your catalog is syncing so your products can appear in ChatGPT discovery.

By November 25

• Freeze structural feed changes. From here on, only update prices and stock.
• Run agent tests: ask Google AI Mode, ChatGPT, Perplexity, and Copilot for your top 10 products and categories using real constraints, and log whether you appear and whether the facts are right.
• Set alerts so out of stock items flip to the correct availability within hours.

November 27 to 30

• Check daily for price mismatches and new disapprovals.
• Update availability fast; nothing loses an agent’s confidence like a sold out item marked in stock.
• Track AI referral traffic and conversions separately in analytics.

From December 1: start the Q1 2027 list

• If you’re a US retailer, review UCP checkout requirements and submit the interest form.
• Set up or expand an ACP product feed for ChatGPT, directly or through a partner.
• Extend conversational attributes across the full catalog.
• Build a monthly agent testing routine.

Feed Health KPIs to Track

These targets are YuvGro’s working benchmarks, not platform rules. Adjust them to your catalog.

Mistakes That Get Products Skipped

• Keyword stuffed or promotional titles. They match fewer real requests and read as low quality.
• Vague availability. OpenAI accepts “unknown” as a value, but it doesn’t tell an agent the item is in stock. Send a real status.
• Changing or reused IDs. Agents and platforms rely on stable IDs to connect prices, reviews, and history.
• Mismatched review data. Review counts and ratings must describe the same set of reviews.
• Hidden costs. Shipping fees revealed only at checkout make your offer look cheaper to people and worse to agents once they find out.
• A page that disagrees with the feed. When the two conflict, you lose trust with both the platform and the agent.

How YuvGro Helps?

We run product feed audits and AI shopping visibility programs for ecommerce brands across the US, the UK, India, and our other markets: feed cleanup, conversational attributes, product page markup, agent testing, and holiday readiness. Explore our assistive agent optimization services and AI optimization services. When we publish our deep dive on structured data for AI Mode and search agents, it will cover the markup side in more detail.

The Bottom Line

AI shopping agents can only recommend what they can read and verify. A complete, current, consistent feed is the cheapest, fastest way to show up more often in AI Mode, ChatGPT, and Perplexity, and it’s the foundation every checkout program builds on.

Work the preflight checklist before Black Friday. Start the protocol integrations in January.

Frequently Asked Questions

How do I get my products recommended by ChatGPT?

Share complete, accurate product data with ChatGPT through the Agentic Commerce Protocol, either directly or through a partner such as Salesforce or Stripe; Shopify merchants are included through Shopify Catalog. Include the nine required fields (ID, title, description, URL, brand, seller name, image, availability, and price), add GTINs, variants, shipping, returns, and ratings, and keep prices and stock current. Note that standard uploads currently target the US.

How do I optimize Merchant Center for AI Mode shopping?

Fix disapprovals and price or availability mismatches, write attribute rich titles and factual descriptions, complete identifiers and variants, and add shipping and return settings. Then add conversational attributes such as product Q&A and related products through a supplemental data source, and use the AI performance insights report, where available, to find missing attributes and top shopper terms.

What product attributes do AI shopping agents use?

Agents rely on the core attributes (title, description, price, availability, brand, image, identifiers such as GTIN) plus anything that matches a shopper’s constraints: size, material, compatibility, variants, shipping cost and speed, return terms, and ratings. On Google, conversational attributes such as product Q&A and related products add context that helps match detailed, conversational requests.

What are Merchant Center conversational attributes?

Conversational attributes are optional Merchant Center fields that help AI systems understand product nuances on surfaces like AI Mode. There are six: question and answer, document link, related product, item group title, variant option, and popularity rank. Google recommends submitting them through a supplemental data source, and they don’t affect the approval status of existing products.

Do I need structured data on product pages for AI Mode?

Google says structured data isn’t required to appear in AI Mode or AI Overviews. It’s still worth adding complete Product markup because it supports merchant listing eligibility and helps confirm your facts, as long as every value in the markup matches what’s visible on the page. Store wide return and shipping policies can be defined once at the Organization level.

Can I sell directly inside Google AI Mode?

Yes, if you’re eligible. Google’s UCP powered checkout lets shoppers buy from eligible retailers directly in AI Mode and the Gemini app using Google Pay, while you remain the seller of record. It’s currently available to select merchants with product eligibility in the US, Canada, and Australia, and listings must use the native commerce checkout attribute to show the Buy button.

How often should I update my product feed during Black Friday?

Price and stock changes should reach your feed within hours, not days. During Black Friday and Cyber Monday, increase scheduled fetches or use the Merchant Center API, freeze structural feed changes a few days before, and check daily for mismatches and disapprovals. Mark sold out items correctly as quickly as possible.

Does Perplexity have a product feed?

Perplexity draws heavily on web retrieval, and it has offered merchant and payment partnerships, including Instant Buy with PayPal in the US since November 2025. Check Perplexity’s current merchant terms directly. Either way, crawlable, complete product pages whose prices and details match your feeds are the foundation for being picked.