Search has all the time beloved structured knowledge. For years we’ve shipped well-formed product catalogs to Google Service provider and tuned attributes to win in Buying. That psychological mannequin nonetheless helps—however OpenAI’s new commerce feed is drastically evolving the sport. We’re not optimizing static listings for a SERP; we’re supplying machine-readable context to an assistant that negotiates intent in actual time.
Right here’s the way it’s totally different and the way we’ll method it with semantic knowledge.
What’s essentially totally different
I spent a number of hours tinkering and reviewing the specs and studying a number of of the early commentary. Right here is my take:
1) From “itemizing” to “reasoning enter.”A Google Service provider feed describes a product so it may be ranked and priced in a comparability grid. An OpenAI feed turns into proof the assistant makes use of whereas reasoning inside a dialog. Attributes aren’t merely filters; they’re details the agent mentions, contrasts, and turns into solutions.
2) From key phrases to use-cases.Google expects taxonomy + attributes. OpenAI advantages from use-case semantics: what the product is for, what it pairs with, who it matches, and underneath what constraints (funds, supplies, supply window, accessibility, sustainability). Suppose “suitable with iPhone 15 Professional,” “works for chilly brew,” “matches 28–30″ waists,” “vegan leather-based,” “next-day in Milan.”
3) From click-through to job completion.Buying adverts chase CTR and conversions off-site. Agentic commerce optimizes for drawback decision in-chat: shortlisting, trade-off explanations, bundle ideas, and—if enabled—on the spot checkout. The standard bar shifts from shiny creatives to scrub details, variant logic, stock freshness, and clear insurance policies. Keep in mind, ChatGPT controls its personal reminiscence of you.
4) From one PDP to many variants.Assistants excel once they can purpose throughout variants (dimension, coloration, area, worth, success). Which means embracing item_group_id + per-SKU provides, not flattening every thing into the dad or mum. Variant hygiene is not optionally available. GS1 Digital Hyperlink knowledge is foundational for higher-quality, context-aware product interactions, delivering a particular expertise throughout classes. And the magic doesn’t cease on-line, these experiences can movement naturally into native retail.
5) From periodic uploads to near-real-time fact.Google tolerates slower cadences. Assistants want recent inventory, worth, and supply estimates to keep away from hallucinations and damaged guarantees. Deal with availability and transport as dwell indicators, not static attributes.
What the OpenAI feed desires (virtually)
Identifiers that received’t wobble: secure id (SKU/supply), item_group_id for variants, plus GTIN/MPN when you could have them.
Plain-text readability: descriptions with out HTML; keep away from advertising fluff that may’t be reasoned over.
Wealthy, typed attributes: situation, supplies, dimensions with items, compatibility, care directions, guarantee.
Associated merchandise: equipment, add-ons, and upgrades that improve compatibility, increase capabilities, or enhance efficiency.
Media that proves it: major + further photographs; video or 3D the place helpful.
Business fact: worth with forex, sale home windows, stock amount, preorder/ETA.
Success actuality: transport areas, prices, and supply estimates (ideally per area/warehouse).
Coverage surfaces: returns, privateness, phrases, vendor id. These construct belief and unlock checkout.
For those who additionally allow Prompt Checkout, plan for the server-side bits (checkout session endpoints, order notifications, fee token dealing with). It’s nearer to an integration than a feed.
A KG-first blueprint for product feeds
1) Mannequin merchandise are entities, not rows.Our Data Graph already expresses Product ↔ Provide ↔ Group ↔ Place. We’ll map entities to feed data whereas preserving relationships the assistant can use (model lineage, compatibility, equipment, retailer places).
2) Promote use-case details to first-class fields.We’ll extract and normalize attributes that reply “Will this work for me, right here, now?”—sizes (with techniques), match, allergens, energy requirements, area availability, and suitable units. These grow to be reason-ready snippets the agent can quote.
3) Deal with variants as residents.Every variant will get its personal entity with a shared item_group_id. Mum or dad PDP stays canonical; variants carry offer-level worth, inventory, personalised description and media.
4) Hold the reality recent.We already be a part of, generally, the KG with stock/OMS and pricing to maintain availability, inventory_quantity, and transport guarantees present. Stale knowledge erodes belief; assistants keep in mind. It will grow to be much more attention-grabbing when the native stock as effectively are added to the combination.
5) Make insurance policies specific and linkable.Clear, crawlable URLs for returns, privateness, and phrases—plus vendor title and make contact with—are necessary for assistant-mediated belief and checkout eligibility.
6) Write like we’re speaking to a colleague.The assistant isn’t swayed by superlatives. We write descriptions, product highlights and particulars which are particular, comparable, and testable: supplies, tolerances, requirements, care, and limitations. All of that is performed already at scale utilizing our Content material Era instrument.
The mindset shift
Google Service provider rewards catalogs that rank. OpenAI rewards information that helps—details an assistant can depend on to information somebody to the precise alternative, confidently and shortly. That’s the place structured knowledge and graph pondering change the sport.
With a Data Graph because the supply of fact, we don’t simply listing merchandise—we mannequin the area and the language round it:
Area entity graph: merchandise, provides, manufacturers, classes, compatibility (“works with”), substitutes, bundles, retailer places, return insurance policies—typed and linked with Schema.org/GS1. This provides the agent grounded relationships to purpose over (e.g., “this filter matches the Aeropress; this lens is for Nikon Z-mount; this dimension maps to EU 42”).
Lexical graph: the textual content aspect—chunked PDP copy, specs, care directions, dimension guides, manuals, opinions, and structured Q&A. We connect every chunk to the precise entities, so the assistant can quote exact, verified passages as an alternative of guessing.
The mixture is highly effective: the entity graph tells the agent what’s linked to what; the lexical graph provides the precise phrases and proof to reply follow-ups (“the right way to clear it,” “is it vegan,” “will it arrive by Friday in Milan?”). That is the distinction between a shoppable catalog and an assistant-ready product reminiscence.
Virtually, we ship three issues within the feed (and hold them recent):
Info: identifiers, worth, inventory, variants, dimensions with items, compatibility.
Context: use-case attributes (match, allergens, energy requirements, regional availability) and coverage surfaces (returns, guarantee).
Proof: linked textual content chunks and Q&A snippets the agent can cite verbatim.
OpenAI (and another AI Agent) then has every thing it wants to elucidate trade-offs, suggest bundles, and full duties—not simply rank SKUs. With a semantic layer (a Product Data Graph), you’re able to ship not simply merchandise, however understanding.
Make your catalog agent-ready with WordLift B+. Begin your Product Data Graph at this time.
Agentic Commerce Protocol (ACP): fundamentals, tech & adoption
What are the important thing technical elements of the ACP structure?
Consider ACP as a common language that permits AI brokers, retailers, and fee techniques to speak securely and effectively. It’s an open normal designed for interoperability.
The core elements embody:
Agent & Intent Schema: A standardized format for an AI agent to specific a consumer’s aim, such because the intent to buy a particular product with outlined constraints (e.g., funds, supply time).
Functionality Discovery: A mechanism for retailers to promote their capabilities—what merchandise are of their catalog, present pricing, availability, and transport choices—so brokers can uncover them.
Standardized Checkout API: A constant, predictable movement for an agent to construct an order, get the consumer’s approval, and authorize fee.
Authentication & Consent: A safe system of short-term tokens and consent data that permits an agent to behave on a consumer’s behalf, with a transparent audit path that the consumer can revoke at any time.
Pluggable Fee Layer: An adaptable interface that connects to numerous fee processors (like Stripe) and strategies (corresponding to bank cards or digital wallets).
Put up-Buy Webhooks: Standardized occasion notifications for success processes, together with stock updates, transport confirmations, and returns.
Safety & Auditing: A framework for cryptographically signing key messages and sustaining detailed logs to forestall fraud and resolve disputes.
Is the Agentic Commerce Protocol suitable with blockchain applied sciences?
Sure. Whereas ACP is blockchain-agnostic—which means it doesn’t require a particular ledger know-how—it’s totally suitable with them. Companies can combine blockchain the place it provides distinct worth, corresponding to:
Auditing & Provenance: Utilizing a permissioned ledger (like Hyperledger) to create an immutable audit path for enterprise compliance.
Digital Belongings: Leveraging public chains (like Ethereum) for commerce involving tokenized property or NFTs.
In essence, ACP can function with conventional fee techniques, blockchain-based ones, or a hybrid of the 2.
What are some real-world functions of the Agentic Commerce Protocol in 2025?
As of late 2025, ACP is enabling a brand new wave of frictionless, AI-driven commerce. Excessive-impact functions embody:
Conversational Checkout: Customers can ask a generative AI assistant (like ChatGPT) for a product, and the agent can full all the buy throughout the dialog, as seen in pilots with retailers on Etsy and Shopify.
Automated Private Buying: AI brokers that mechanically reorder family necessities, dynamically choosing the right service provider primarily based on real-time worth, availability, and supply pace.
Complicated Itinerary Reserving: Journey brokers that seamlessly assemble and buy multi-part journeys, reserving flights, inns, and insurance coverage throughout totally different suppliers in a single, unified course of.
B2B Procurement Automation: Enterprise assistants that autonomously negotiate for traditional merchandise (SKUs) and set off buy orders when stock ranges are low.
How does ACP allow AI brokers to create personalised purchasing experiences?
ACP offers the standardized framework for communication and consent that AI brokers have to act on a consumer’s distinctive preferences. It permits an agent to:
Perceive Person Constraints: Cross vital data like funds, model preferences, or dietary restrictions to a service provider system.
Question Service provider Capabilities: Ask for particular product variants, sizes, or supply home windows.
Execute Tailor-made Purchases: Leverage wealthy metadata from retailers—corresponding to product bundles or customized success choices—to create a hyper-personalized choice and checkout movement for the consumer.
This eliminates the necessity for customized, one-off integrations for every service provider, permitting any agent to offer a customized expertise with any ACP-enabled enterprise.
What are the first enterprise advantages of implementing ACP?
For companies, adopting ACP provides a big aggressive benefit by making their services and products “agent-native.” Key advantages embody:
Elevated Income: Scale back friction and cart abandonment by enabling prospects to make purchases immediately inside AI-powered conversations.
Broader Discovery: Place your merchandise to be really useful by AI brokers, opening up a strong new discovery and gross sales channel.
Future-Proofing: Grow to be an early adopter of the brand new normal for AI-driven commerce, making certain your online business is prepared for the following wave of digital interplay.
Lowered Integration Prices: A single implementation permits you to interoperate with a rising ecosystem of AI brokers and platforms, minimizing the necessity for pricey, bespoke engineering work.
Standardized Safety: Leverage established fee companions (like Stripe) for safe, dependable fee processing and legal responsibility dealing with.
What are the important thing challenges or concerns when adopting ACP?
Adopting any new protocol requires strategic planning. Key concerns embody:
Techniques Integration: Mapping legacy catalog, stock, and success techniques to the standardized ACP knowledge schema.
Operational Readiness: Making ready to deal with new kinds of requests from brokers, corresponding to complicated success logic or multi-part shipments.
Buyer Expertise: Designing a seamless consumer expertise the place prospects really feel in management and perceive how agent suggestions are generated.
Compliance & Regulation: Navigating the evolving authorized panorama for AI, funds, and shopper knowledge safety.
Fraud Prevention: Implementing clear protocols for dealing with disputes and stopping fraud in agent-mediated transactions.
What’s the market adoption fee of ACP in late 2025?
As of late 2025, ACP is in its preliminary industrial rollout part. The protocol was launched by OpenAI and Stripe with the launch of “Prompt Checkout” in ChatGPT, that includes Etsy as the primary dwell service provider companion and Shopify integration introduced to comply with.
Whereas a worldwide adoption proportion has not but been revealed, the open-sourcing of the protocol is predicted to speed up adoption considerably throughout the e-commerce and funds ecosystem.
How does ACP differ from a standard e-commerce platform?
The important thing distinction is that ACP is a protocol, not a platform.
A conventional e-commerce platform (like Shopify or Magento) is a hosted storefront designed for human interplay through a graphical consumer interface (GUI).
ACP is an interoperability protocol that defines an ordinary for programmatic commerce, permitting autonomous AI brokers to find merchandise and transact on a consumer’s behalf.
It’s designed for machine-to-machine communication, with an emphasis on consent tokens, agent intent, and machine-readable catalogs, reasonably than human-driven clicks.
How can ACP be used to optimize provide chain administration?
ACP offers standardized, real-time indicators about product availability, lead instances, and success choices. This permits clever brokers to automate and optimize provide chain operations by:
Automating Reorders: Triggering buy orders mechanically when stock hits a predetermined threshold.
Dynamic Sourcing: Querying a number of suppliers to seek out the best choice primarily based on present inventory ranges and supply instances.
Simply-in-Time Ordering: Enabling agent-driven intelligence to cut back stockouts and reduce carrying prices.
When built-in with present enterprise useful resource planning (ERP) techniques, ACP can considerably enhance provide chain effectivity and resilience.
What are the potential authorized and regulatory implications of agentic commerce?
As of 2025, regulators globally are actively growing frameworks for AI and commerce. Key areas for companies to watch embody:
Client Safety: Defining legal responsibility and return insurance policies for purchases made by an agent on a consumer’s behalf.
Fee Regulation: Making certain compliance with KYC (Know Your Buyer) and AML (Anti-Cash Laundering) guidelines for programmatic transactions.
Information Privateness: Adhering to rules like GDPR and CCPA relating to consent data, knowledge portability, and the precise to deletion.
Promoting & Disclosure: Sustaining transparency in how AI brokers make suggestions to keep away from potential conflicts of curiosity.
Implementers ought to prioritize constructing specific consumer consent flows and sustaining sturdy, auditable data.
How does the Agentic Commerce Protocol deal with consumer knowledge privateness?
The ACP framework is constructed on core ideas of knowledge privateness and consumer management:
Specific Consent: Brokers can solely act on a consumer’s behalf with an specific, revocable consent token.
Scoped Permissions: Agent actions are restricted to licensed scopes, stopping overreach.
Information Minimization: The protocol encourages brokers to request solely the info fields needed to finish a transaction.
Auditability: Customers can evaluation a transparent log of their agent’s actions, making certain transparency.
Whereas the precise implementation relies on the distributors concerned, the protocol’s design prioritizes consumer privateness.
What industries are most certainly to learn from ACP adoption?
ACP is poised so as to add important worth to industries characterised by complicated purchases, repeat orders, or the necessity for dynamic discovery.
Early adopters are prone to be in:
Retail & E-commerce: Notably marketplaces and direct-to-consumer (DTC) manufacturers.
Journey & Hospitality: For assembling complicated, multi-part bookings.
B2B Procurement: For automating repeat orders and sourcing normal SKUs.
Subscription Companies: For managing automated reorders of groceries, medicines, and different consumables.
Digital Items: Together with marketplaces for NFTs or different tokenized property the place agent-driven discovery can speed up transactions.
What are the newest main bulletins relating to ACP in 2025?
In late September 2025, OpenAI, in partnership with Stripe, formally launched the Agentic Commerce Protocol (ACP). This launch coincided with the discharge of “Prompt Checkout” inside ChatGPT, a characteristic that permits customers to finish purchases conversationally.
Prompt checkout, ChatGPT + Stripe, safety, comparability
What’s Prompt Checkout, and the way does it work in ChatGPT?
Prompt Checkout is a streamlined buy movement designed to reduce friction. By leveraging securely saved fee, transport, and consent knowledge, it permits a buyer to finish a purchase order with a single affirmation, eliminating the necessity to fill out conventional checkout varieties.
Within the context of an AI assistant like ChatGPT, this course of is much more seamless. The AI agent constructs the acquisition order immediately throughout the dialog, presents the phrases, and upon consumer affirmation, authorizes the transaction through a fee processor like Stripe. This permits a purchase order to occur on the level of discovery, making a frictionless, in-context shopping for expertise.
What’s the Agentic Commerce Protocol (ACP)?
The Agentic Commerce Protocol (ACP) is an open normal, co-developed by OpenAI and Stripe, that standardizes how AI brokers, retailers, and fee techniques work together to programmatically uncover merchandise and full purchases.
It capabilities by defining machine-readable schemas and API endpoints that permit an agent to execute all the commerce lifecycle: discovering a service provider’s merchandise, establishing an order, requesting consumer consent, authorizing fee, and receiving success updates—all by way of a safe and standardized messaging framework.
How does ACP differ from conventional e-commerce APIs?
The first distinction is that ACP is designed particularly for autonomous and conversational brokers, whereas conventional e-commerce APIs are constructed for human-driven storefronts.
Agent-Centric Design: ACP is constructed round ideas like agent intent and consent tokens, which aren’t central to conventional APIs.
Interoperability: It goals to create a common normal, permitting any agent to work together with any ACP-enabled service provider, thus lowering the necessity for pricey, customized point-to-point integrations.
Standardized Orchestration: ACP codifies all the agent-led fee and post-purchase movement in a approach that standard e-commerce APIs don’t.
What applied sciences allow Prompt Checkout and Agentic Commerce?
This new commerce paradigm is made potential by the convergence of a number of key applied sciences:
Massive Language Fashions & Agent Platforms (e.g., ChatGPT)
Standardized Protocols (i.e., the Agentic Commerce Protocol)
Superior Fee Processors (e.g., Stripe)
Digital Pockets Requirements (e.g., Apple Pay, Google Pay)
Trendy Authentication & Telemetry Layers (e.g., OAuth, JWTs)
How do funds work with Prompt Checkout?
Stripe serves because the core fee infrastructure for the preliminary rollout of Prompt Checkout in ChatGPT. It offers a connector that implements the ACP checkout movement, permitting the AI agent to request fee authorization securely.
At launch, this integration helps the frequent fee strategies that Stripe already processes, together with main credit score/debit playing cards and digital wallets like Apple Pay and Google Pay.
What are the advantages of Prompt Checkout for customers and retailers?
For customers, the first advantages are:
Velocity and Comfort: The flexibility to buy a product immediately inside a dialog with out filling out varieties.
Contextual Relevance: Purchases may be made on the precise second of discovery.
For retailers, this interprets on to:
Greater Conversion Charges: The numerous discount in friction results in decrease cart abandonment.
How can retailers allow Prompt Checkout in ChatGPT?
To grow to be “agent-native,” retailers have to:
Expose ACP-compatible API endpoints for his or her catalog, product availability, and success choices.
Combine a supported fee processor (Stripe is the preliminary companion).
The open-source ACP sources and documentation from Stripe are designed to make this integration course of easy for retailers.
What are the present limitations of Prompt Checkout (as of September 2025)?
As a brand new know-how, the preliminary rollout has some limitations:
Geographic & Service provider Availability: The launch is presently U.S.-only, with a restricted variety of collaborating retailers (Etsy is dwell, with Shopify integration deliberate).
Single-Merchandise Purchases: The preliminary implementation is for single-item checkouts, with multi-item cart performance deliberate for future enlargement.
Geographic & Service provider Availability: The launch is presently U.S.-only, with a restricted variety of collaborating retailers (Etsy is dwell, with Shopify integration deliberate).
Single-Merchandise Purchases: The preliminary implementation is for single-item checkouts, with multi-item cart performance deliberate for future enlargement.
