For the reason that ChatGPT second, on November 30, 2022, it was clear that what we now name AI (LLMs) would have slowly then immediately redefined the entire search engine business.
Effectively, nearly three years into that journey, we’re getting there. In actual fact, if within the final three years, LLMs have redefined a very good chunk of the informational journey for search, and discovery, we’re now coming into one other important part: the redefinition of the transactional journey!
What you’re witnessing isn’t simply one other e-commerce interface—it’s the demise of the “Go to” button.
From Discovery to Autonomous Transaction
An instance of a immediate inside ChatGPT, which, due to the semantic enrichment, allows Ray-Ban to get visibility throughout AI search, but that is solely the beginning of a journey that’s shifting from discovery to a complete transactional journey the place the agent aligns with the human and completes the acquisition.
That Ray-Ban buying expertise in ChatGPT?
It’s the final technology of AI-assisted commerce. The subsequent technology? The AI completes the acquisition for you, no redirect required.
Immediately’s journey is already out of date. You ask ChatGPT about sun shades, it surfaces Ray-Ban choices with costs, and you then click on “Go to Ray-Ban” to finish the acquisition on their web site.
Tomorrow’s actuality is being constructed proper now. You ask ChatGPT about sun shades, it surfaces Ray-Ban choices with costs, you say “Purchase these for me,” and the transaction completes. Sun shades arrive in 2 days.
No web site. No redirect. No checkout type.
WordLift’s Agentic Commerce 5-Layer Semantic Stack
WordLift framework for agentic commerce.
This isn’t magic—it’s structure. Right here’s the technical basis making autonomous commerce potential:
Layer 1: Entity/Supply (The Product Feed)
Your catalog turns into AI-readable. GTIN codes, variant buildings, stock standing, success choices—all mapped to standardized schemas that brokers can parse and purpose over. If Ray-Ban’s sun shades aren’t correctly structured right here, they don’t exist to the AI.
Layer 2: State/Stream (The Checkout Spec)
The AI manages your cart state. Canonical endpoints expose line gadgets, totals, delivery choices, and order standing. The agent doesn’t simply learn your catalog—it might modify cart contents, apply reductions, and progress by means of checkout states identical to a human would.
Layer 3: Belief/Authorization (Delegated Cost)
Safe tokenized transactions with out exposing card information. Constructed on protocols like OpenAI‘s Agentic Commerce Protocol (ACP) with Stripe, this layer lets customers authorize spending limits and expiry home windows. The agent can spend in your behalf, however inside boundaries you management.
Layer 4: Interplay/UI (Apps SDK/MCP)
How the agent renders and causes. This layer separates structured information (what the mannequin thinks about) from presentation information (what you see). The agent understands “Ray-Ban Wayfarer, $139, in inventory” internally, then renders it conversationally: “I discovered these basic Wayfarers for you.”
Layer 5: Discovery (AEO Floor)
How manufacturers get chosen within the first place. That is website positioning 3.0: metadata, prior utilization patterns, context matching, and linking state. In case you’re not optimized right here, you by no means enter the consideration set. The perfect product feed on this planet is nugatory if the agent by no means thinks to question you.
The WordLift Bridge: Concept Meets Observe
WordLift routinely transforms your current catalog into OpenAI-compliant schemas.
Two important toggles management every little thing: enable_search makes your merchandise discoverable in ChatGPT, whereas enable_checkout permits direct buy completion with out leaving the chat.
Behind the scenes, it generates the OpenAI Commerce JSON, which maps your stock, pricing, variants, and success right into a semantic layer that brokers can perceive.
That is the infrastructure play. Whereas manufacturers debate which AI search instrument to purchase, the good ones are rebuilding their information foundations.
The Core Perception
“Most manufacturers are nonetheless optimizing floor content material. Most manufacturers are nonetheless determining in the event that they want Profound, OtterlyAI or Scrunch. The fact is that the actual bottleneck is deeper: your information form.“
Translation: You possibly can’t prompt-engineer your approach into an agent’s buying cart in case your product information isn’t semantically structured.
The outdated bottleneck was website positioning rankings, advert spend, and conversion charge optimization. The brand new bottleneck is information legibility, schema compliance, and API readiness. In case your catalog isn’t machine-readable in the way in which brokers count on, you don’t exist. No quantity of promoting finances fixes that.
What Modifications (And What Doesn’t)
The “Go to” button dies. Redirects to exterior websites change into friction. Web site site visitors as a KPI turns into meaningless when brokers don’t browse, they transact. Conventional conversion funnels collapse when buy occurs in a single message. Web page-based website positioning issues much less when ChatGPT by no means sends customers to your web site.
Belief and authorization survive. Customers nonetheless management spending limits and agent permissions. Bodily success stays unchanged—somebody nonetheless ships the sun shades. Model differentiation nonetheless issues as a result of the agent wants causes to select Ray-Ban over Oakley. Customer support endures as a result of returns, exchanges, and help nonetheless require human interplay.
One thing new emerges: Semantic website positioning (website positioning 3.0). Manufacturers optimize for agent discoverability, not web page rankings. That is invocation over indexation—manufacturers aren’t discovered, they’re known as by brokers. Knowledge form turns into a aggressive moat as a result of the best-structured catalog wins. Product design shifts to agent-first considering, with descriptions, variants, and attributes optimized for AI reasoning somewhat than human searching.
What’s Subsequent for Agentic Commerce?
The specs are public. The requirements are open. OpenAI’s Product Feed Specification is stay and documented. Agentic Commerce Protocol (ACP) is constructed with Stripe and open to any service provider. Mannequin Context Protocol (MCP) supplies the muse for Apps SDK integration.
What’s lacking? Your information mapped to them.
Don’t begin by shopping for extra AI search trackers or optimizing prompts with out structured information. Don’t construct customized LLM interfaces earlier than fixing your backend. That’s constructing a penthouse on quicksand.
Begin right here: Audit your product information. Do you may have clear GTIN/MPN codes? Structured variants? Actual-time stock standing? If not, you’re not prepared for agentic commerce. Then map to OpenAI’s Product Feed Spec utilizing instruments like WordLift’s Feed Transformer or customized pipelines. Your catalog should communicate the semantic language brokers perceive.
Subsequent, implement checkout state endpoints that expose canonical APIs for creating, modifying, and finishing a cart. The agent must handle the state, not simply learn it. Allow delegated fee infrastructure by integrating with ACP-compatible fee processors. Tokenized, constrained authorization is non-negotiable.
Lastly, optimize for discovery (AEO) by means of metadata, categorization, and context matching. If the agent doesn’t consider you, your good product feed is nugatory.
