Key Takeaways
- Agentic commerce readiness is a product data problem, not a checkout project: agents shortlist what they can read and match to a shopper's question.
- Google's conversational attributes are optional and do not affect approval, which makes them an early-mover window rather than a requirement.
- Work in three layers: solid identifiers and core attributes, then conversational fields, then data that stays fresh.
- Skip the speculation about checkout protocols. The payment layer will reach you through your platform; the data layer is yours to fix now.
- Start with your best sellers, measure before and after, and extend to the long tail once the process works.
Most advice about agentic commerce readiness starts with checkout: new protocols, wallets, payment tokens. That is the loud part of the story, and it is mostly out of your hands. The quiet part is the one you can act on this week. Before an AI agent can buy anything from you, it has to find your product, understand it and decide it fits the shopper’s request. All of that happens in your product data.
If you want the definition first, What Is Agentic Commerce? covers the basics. This post is the practical follow-up: what to fix, in what order, and what to leave alone while the standards settle.
What Agentic Commerce Readiness Actually Means
Readiness is not a badge you earn or a switch you flip. It is how much of your catalog an AI system can use without guessing.
Picture a shopper asking an assistant for “a waterproof jacket for wide shoulders, under $150, that ships this week.” The agent does not browse your site. It reads structured records and checks each condition: waterproof, fit, price, delivery. A product whose record states all four is easy to include. A product whose record says only “Men’s Rain Jacket, Blue” leaves three conditions unanswered, so the agent either skips it or has to infer, and inference is exactly what these systems are built to avoid.
So readiness is really a question of coverage. How many of your products answer the questions shoppers are about to ask? For most stores the honest answer is “fewer than we thought,” because the default export from a store platform was built for ads and listings, not for conversation.
Why Optional Fields Are the Early-Mover Window
Here is the part merchants tend to miss. Google’s newer AI-facing fields, the conversational attributes introduced at Google Marketing Live 2026, are optional. Google says they do not affect product approval. Leave them empty and nothing breaks: no disapproval, no warning, no drop in your existing Shopping listings.
That sounds like a reason to ignore them. It is the opposite. Fields that carry no penalty and no reminder get filled by the few merchants who understand what they are for, and skipped by everyone else. The gap between those two groups is the opportunity. When AI Mode or another AI surface has to pick between two products that sell the same item, the one with structured highlights, specs and answered questions gives it more to work with.
Optional fields have a habit of becoming expected ones. Nobody can say when, or whether, that happens here, and it would be wrong to promise a ranking benefit. But the cost of preparing is low, the work carries over to every surface that reads structured product data, and the merchants who wait until it is required will be doing it under pressure. For a field-by-field breakdown, see The 8 Conversational Attributes for AI Shopping Visibility.
The Three Layers of Agentic Commerce Readiness
Think of readiness as a stack. Each layer depends on the one beneath it, so work from the bottom up.
Layer 1: A foundation of complete, honest basics
Before anything conversational, the basics have to be right. That means accurate titles that lead with brand and product type, real identifiers, correct prices and availability, and the core attributes shoppers actually filter on: color, size, material, gender and age group where they apply.
Identifiers deserve a special warning. If a product lacks a GTIN, the fix is to get the real one from the manufacturer or GS1. Never invent one. A made-up code belongs to someone else’s product or to nothing, and it will cause far more trouble than a gap. A good audit flags which products lack identifiers so you know where to go and ask.
Gaps at this layer are also the ones that cost you today, in ads and listings, regardless of AI. How Google Punishes Missing Feed Attributes shows how those gaps escalate from limited performance to silent exclusion.
Layer 2: Conversational fields that mirror real questions
This is the new layer. Google documents fields built for AI surfaces: product highlights (short benefit statements), product details (structured specs), question and answer pairs, and variant options that explain how items in a family differ. The question and answer field is described as intended for conversational experiences such as AI Mode.
The principle that makes these fields work is simple. Write them in the shape of the questions shoppers ask. “Is it waterproof?” becomes a question and answer pair with a specific answer. “What is it made of?” becomes a structured detail row. A highlight states a benefit in plain words, not a spec. The rules for length and wording matter, and breaking them wastes the effort, so use How to Write product_highlight, product_detail and question_and_answer as your reference.
One more rule applies across the layer: only write what is true of the product. These fields are not a place for health claims or invented benefits. Accurate and specific beats impressive and vague every time, for the agent and for the shopper who clicks through.
Layer 3: Freshness you can trust
An agent that recommends a product is making a promise on your behalf, so it cares whether the price and stock it read are still true. A feed that updates once a week and drifts away from your site is a quiet liability. Mismatches between feed and page are among the first things to pull a product off the shelf.
You do not need to chase real-time perfection on day one. You need a feed whose update rhythm matches how fast your catalog actually changes, and a check that catches price and availability drift before a platform does.
Optional fields do not stay optional forever. The merchants who treat a no-penalty window as a reason to wait are the ones who end up doing the work under a deadline.
What Not to Spend Time On
A few things look like readiness work and are not.
Chasing the checkout protocols. Standards for agent payments are being built by the large platforms and payment networks. You will adopt them through your store platform and payment provider when they are ready. Reading every spec update will not improve a single product record.
Rewriting your site copy first. Page copy still matters for people who land on your store. But an agent reads the structured record, and a polished product page sitting on top of a thin feed answers none of its questions.
Paying for placement promises. Nobody can guarantee recommendation inside an assistant, and anyone who does is selling something. Spend the effort on the data you control.
A 30-Day Readiness Plan
You can do this without a rebuild. Here is a sequence that works for most stores.
- Week one: audit. Pull your feed into a spreadsheet or a scoring tool and measure completeness per field. Note which products lack identifiers, which attributes are empty on more than a third of the catalog, and which titles are thin.
- Week two: fix the foundation. Correct titles, fill missing core attributes where the information exists in your own data, and source real identifiers for the products that lack them.
- Week three: add conversational fields to your top sellers. Start with highlights, details and a few question and answer pairs for the products that earn the most. This is your test group.
- Week four: measure and extend. Compare completeness before and after, check Merchant Center diagnostics for new issues, and roll the process out to the next tier of products.
Doing step three by hand across thousands of SKUs is where most teams stall. It is also where a supplemental feed helps: it sits on top of your primary feed and adds the missing attributes without touching the original. Why Supplemental Feeds Matter More as AI Shopping Grows explains how that layer works and why it survives platform updates.
Measure as you go. A score per product lets you sort your catalog by weakness instead of guessing, and a before and after comparison shows whether the work moved anything. What Is a UCP Score? explains the idea in plain terms.
Conclusion
Agentic commerce will arrive unevenly, with the payment and checkout pieces settling last. The data piece is different: it is available to work on now, it carries no penalty for starting early, and it pays off on every surface that reads structured product information, from Google Shopping to AI assistants. Fix the basics, add the conversational layer, keep the feed fresh, and measure the change.
If you would rather not hand-write hundreds of highlights and answers, UCP Radar connects to your Google Merchant Center account, checks your feed against more than 50 validation rules, scores each product, and generates the optimized supplemental feed for you to register in Merchant Center. You stay in control of what gets submitted. The agents are still learning how to shop; your job is to make sure there is something worth finding when they do.
Frequently Asked Questions
Agentic commerce readiness is how well your product data can be read, matched and cited by an AI shopping agent. It covers complete identifiers and attributes, the newer conversational fields Google has added for AI surfaces, and data that stays current. It is a measure of your catalog, not of your checkout.
No. Google states that conversational attributes do not affect whether a product is approved. They are optional fields that feed AI surfaces such as AI Mode. That is exactly why they are an early-mover opportunity: nothing forces you to fill them, so most feeds still leave them empty.
Not as the first step. Checkout and payment standards are being built by the platforms and payment providers, and you will adopt them through the tools you already use. What you control today, and what decides whether an agent shortlists you, is the quality of your product data.
The fixes themselves can be quick, but doing them by hand across a large catalog is slow. A realistic plan is to audit and fix identifiers and core attributes in the first week, then add conversational fields to your best sellers first, and extend to the long tail once you see the process works.
No. Nobody can promise placement inside an AI assistant. Readiness removes the reasons a product gets skipped, such as missing attributes and unclear specs, so that you are eligible and easy to understand when an agent goes looking. The assistant still makes the final call.
