Agentic Commerce: Is Your Catalog Ready for AI Buying Agents?
Summarize this article with AI
What is agentic commerce?
3 to 5 trillion dollars in retail spending redirected by AI agents by 2030, according to Gartner. A trend? No. A strategic imperative for every e-commerce brand.
Agentic commerce describes a model where autonomous programs — AI buying agents — search, compare, negotiate, and purchase products on behalf of a consumer. The user defines a need (« a waterproof coat, size M, delivered within 48 hours, max budget 150 euros ») and the agent handles the entire journey.
What once belonged to science fiction is now operational:
- ChatGPT Shopping is live on Etsy, Amazon, and hundreds of marketplaces — the user describes what they’re looking for, the agent displays product recommendations with prices and reviews
- Microsoft Copilot Checkout integrates Shopify, PayPal, and Stripe to complete the purchase directly in the conversational interface
- Perplexity Shopping offers an augmented comparator that synthesizes product sheets, reviews, and shipping conditions into a single response
According to a Shopify Enterprise study (2026), the majority of e-commerce catalogs remain invisible to these agents. Why? Product sheets are designed for human eyes, not for reasoning algorithms.
Why 73% of buyers already use AI in their purchase journey?
Adoption is already massive. 73% of online buyers use at least one generative AI tool during their shopping journey (Salesforce, State of Commerce 2026).
Usage breaks down like this:
- 45% to generate product ideas (« What gift for a beginner hiker? »)
- 37% to get summaries of customer reviews synthesized in one sentence
- 32% to compare prices across multiple retailers
- 28% to check availability and shipping times
Zero-click commerce: purchase directly in the LLM
The model is shifting toward zero-click commerce: the purchase is completed in the agent’s interface. No visit to the merchant’s site. Copilot Checkout illustrates this: product selection, size choice, payment, confirmation. All in 3 conversational exchanges.
This paradigm shift has a direct consequence: 70% of consumers report being comfortable with an agent completing a purchase on their behalf (McKinsey, Consumer Pulse Q1 2026). Trust in transactional AI is established.
For brands, the question is clear: will your catalog be the one the agent recommends, or the one it ignores?
How do AI agents read your catalog?
An AI buying agent doesn’t work like Google. It reasons instead of indexing. To reason, it needs structured, explicit, complete data.
Schema.org: the shared language of agents
Agents rely on Schema.org markup to understand your product sheets. Three essential schema types:
- Product: name, description, brand, SKU, image, category
- Offer: price, currency, availability, shipping conditions, stock status
- AggregateRating: average rating, review count, rating distribution
An agent matches this data against the user’s request. If your sheet is missing an attribute — material, delivery time — the agent favors a competitor who provides that information.
Structured sheets vs. free text
The most common pattern on e-commerce catalogs: rich descriptions, poor structured data. An AI agent prefers an attribute « material: organic cotton certified GOTS » over a 200-word paragraph that mentions cotton in passing.
Agent-Readiness Score in 5 axes
We’ve built a proprietary évaluation framework: the Agent-Readiness Score. It measures a catalog’s capacity to be read, understood, and recommended by an AI agent. 5 axes:
- Intelligibility (20 pts): clarity of HTML structure, heading hierarchy, semantic markup
- Product attributes (25 pts): completeness of normalized attributes (size, color, material, weight)
- Product information (20 pts): unique description, FAQ, sizing guides, technical specs
- Categorization (15 pts): logical hierarchy, breadcrumbs, ProductGroup
- Transactional data (20 pts): price, stock, shipping, returns, payment methods in Offer schema
The majority of French e-commerce sites exploit less than half the potential of their structured data, according to Shopify and Search Engine Land (2026). The opportunity is real for brands that structure their catalog first.
What are the 7 actions to become agent-ready?
Each action targets an axis of the Agent-Readiness Score. Applied together, they transform a passive catalog into one that AI agents actively recommend.
Each product sheet must include a Product markup with at minimum: name, description, image, brand, sku, gtin (EAN/UPC), category. AI agents use these fields to identify your product uniquely. A product with GTIN is identified 3.4x faster than one with a name alone.
Size, color, material, weight, dimensions: these attributes must appear in additionalProperty markup (PropertyValue). Use normalized values (ISO sizes, standard colors, SI units). An agent comparing products needs homogeneous data across sources.
Each product sheet should include 3 to 5 questions-answers marked up as FAQPage. Agents extract these FAQs to answer user-specific questions: « Is this coat suitable for heavy rain? ». On our optimized sites, product FAQs generate +27% rich impressions.
Agents weight social proof heavily. An AggregateRating markup with ratingValue, reviewCount, and bestRating lets the agent instantly compare your product to competitors. Individual reviews in Review schema enrich the synthesis the agent presents to the user.
The Offer markup is critical for transactions. Include: price, priceCurrency, availability (InStock/OutOfStock), deliveryLeadTime, shippingDetails, hasMerchantReturnPolicy. An agent with all transactional data can recommend your product with confidence.
Multimodal agents (GPT-4o, Gemini) analyze images. Use high-definition images (min. 1200px wide) with alt text that describes the product precisely: « Men’s waterproof coat navy blue – front view – adjustable hood ». Each variant (color, angle) deserves its own marked-up image.
The llms.txt file at your site root tells LLMs which pages are priority and how to interpret your catalog. Complete it with a dedicated XML sitemap listing your product sheets with lastmod and priority. Agents respect these directives to prioritize their crawl.
Example llms.txt structure:
> Catalog of 2,400 outdoor products
## Product sheets
– /products/: complete index
– /sitemap-products.xml: products sitemap
## Main catégories
– /category/coats/
– /category/shoes/
– /category/accessories/
What’s the impact of the Google March 2026 Core Update on agentic commerce?
March 2026 reshuffled the deck. The valued signals align exactly with AI agent criteria: original data, proven expertise, fast user expérience.
Winners: original data and expertise
Sites publishing original data — product tests, benchmarks, proprietary photos — gained an average +22% visibility on commercial queries. AI agents favor sources with verifiable expertise. Exactly what Google rewards now.
Losers: affiliation and generic content
Affiliate sites with no added value lost –71% organic visibility. Content AI-generated and published as-is is systematically demoted. Only expert content enriched with proprietary data keeps its positions.
Technical performance: LCP critical
The update strengthens Core Web Vitals weight. Sites with LCP above 3 seconds lost an average –23% traffic. For an AI agent, crawl speed counts double: a slow site is expensive to analyze, so it’s deprioritized.
The overlap is striking. What Google values in 2026, AI agents already require. Optimizing for agentic commerce is optimizing for Google. Same battle.
