Google UCP & OpenAI ACP: 2026 expert checklist to be ready for agentic commerce

Clickless commerce is no longer a theoretical concept. With the Universal Commerce Protocol (UCP), Google is structuring its response to the Agentic Commerce Protocol (ACP) introduced by OpenAI. Two protocols, two ecosystems, but one shared disruption: AI agents now discover, compare, and purchase products on behalf of users.

We have already explored these shifts in two in-depth articles:

• Google Universal Commerce Protocol (UCP): understanding the clickless commerce revolution and its impact on e-commerce
• OpenAI launches ACP: when ChatGPT becomes a sales channel

In this guide, the goal is no longer to understand—but to act. We offer a structured, decision-oriented and implementation-focused framework to help e-commerce, product, data, and marketing teams prepare concretely for the rise of agentic commerce.

What Google and OpenAI are officially saying

Google UCP: product data as a single point of integration

According to Google’s official documentation, UCP aims to eliminate fragmented technical integrations by offering a single entry point into Google’s agentic interfaces: AI Mode in Search, Gemini, Business Agent, and, soon, other conversational surfaces.

Three major implications emerge:

• The Google Merchant Center feed becomes the primary source through which AI interprets a merchant’s offer.
• The quality and structure of product attributes (product data, shipping, returns, availability) directly determine eligibility for agentic purchase journeys.
• The merchant remains the seller of record, but the interface and purchase journey are operated by AI.

In other words, the e-commerce site is no longer the main entry point to commerce; product data becomes the shared language between merchants and agents.

OpenAI ACP: towards a conversational standard of commerce

Even though OpenAI has not yet published documentation as formalized as Google’s, the strategic trajectory is clear. With Instant Checkout, ChatGPT becomes a native transactional channel, capable of recommending products, addressing objections, and completing purchases without redirection.

This scenario is all the more credible because advertising is gradually coming to ChatGPT. As soon as a conversational environment becomes both a media channel and a transactional channel, OpenAI will have to have a structured, comparable and controllable product repository — just like what Google Merchant Center represents in the Google ecosystem.

To function effectively, an agent like ChatGPT cannot rely on:

• heterogeneous web pages,
• implicit or inconsistent descriptions
• information that is difficult to compare between merchants.

On the contrary, it needs standardized product data, usable automatically, capable of feeding both recommendation, comparison and advertising distribution.

The convergence is therefore obvious: like Merchant Center for Google, a product flow standard is a logical and inevitable evolution in the ACP ecosystem. Merchants who have already structured and enriched their product data will be in the best position to quickly activate these new formats, without technological breakthroughs or late catch up.

A shared reality: everything depends on product data maturity

UCP and ACP converge on the same strategic truth: the future of agentic commerce does not depend on the platform you choose, but on the maturity of your product data.

Regardless of the agent involved—Gemini, ChatGPT, or a third‑party agent—the prerequisites are identical:

• a rich and exhaustive feed,
• clear and consistent structure,
• continuous updates,
• a strong focus on machine understanding.

This is why optimizing your Google Merchant Center feed today already constitutes a direct preparation for UCP and an indirect preparation for ACP.

Implementation guide: the 9 key levers to become agentic‑ready

1. Map your real exposure to agentic commerce

Before taking action, it is critical to understand where and how agentic commerce will most impact your business. Not all categories, catalogs, or business models are equally exposed to AI agents. Some products are naturally suited to automated recommendation, while others still require strong branding, guidance, or on‑site storytelling.

What to analyze:

• dependency on Shopping and Performance Max campaigns,
• share of revenue driven by high‑intent queries,
• product decision complexity (simple vs comparative),
• existing e‑commerce, data, and PIM stack.

What to validate:

• which products can already be recommended frictionlessly by AI,
• which categories still require a strong role for the site and brand narrative

2. Building a single source of truth product

Agentic commerce relies on a simple premise: AI can only recommend what it understands unambiguously. Fragmented data, poorly modeled variants, or inconsistent rules create uncertainty for agents—and therefore loss of visibility.

What to implement:

• a stable, shared product identifier (SKU / ID),
• clear and exhaustive variant modeling,
• a coherent product hierarchy (categories, use cases),
• real‑time synchronization of prices, stock, and availability.

Without this foundation, neither UCP nor ACP can be effectively leveraged.

3. Transforming product flow into a real commercial API

Product feeds are no longer just media activation assets. In an agentic environment, they become the raw material directly consumed by AI agents to understand, compare, and recommend products. With UCP, Google already relies heavily on the Merchant Center feed to power AI Mode, Gemini, and agentic transaction flows. Tomorrow, the same logic will apply on the OpenAI side with ACP. A partially filled, poorly structured, or static feed mechanically limits commercial visibility.

Conversely, a rich, coherent, continuously updated feed becomes a structural competitive advantage, immediately activatable and durable over time.

What to do concretely:

• audit the real quality of your GMC feed (completeness, consistency, freshness),
• identify missing or ambiguous critical attributes (use cases, materials, compatibilities),
• enrich titles and descriptions for machine understanding—not just SEO,
• strictly normalize variants to eliminate AI ambiguity,
• align feeds, business rules, and product pages,
• implement frequent and automated updates,
• industrialize enrichment at scale.

In this context, investing in a dedicated solution like Feed Enrich is no longer a tactical choice but a structural decision. Powered by specialized AI models and a proprietary scraper, Feed Enrich enriches, optimizes, and transforms Google Merchant Center feeds into true AI‑ready assets—both performant for Google Shopping campaigns and capable of powering current Google use cases (UCP, Gemini), while seamlessly preparing for future ACP standards on the OpenAI side.

Recommended additional content: The complete guide to optimizing Google Merchant Center product feeds

4. Prepare the site to be read, understood, and used by AI agents

Even though the role of the e‑commerce site is being redefined, this does not mean it will disappear. The site remains a fundamental source of truth for AI agents, which analyze it to understand the offer, validate data consistency, and assess merchant credibility. Tomorrow’s site is no longer designed solely for human visitors. It must also be readable by bots, AI agents, and LLMs, capable of extracting structured data to feed reasoning and recommendations.

An unstructured, inconsistent, or machine‑poor site becomes invisible—even if the offer itself is competitive.

What to do concretely:

• deploy comprehensive schema.org markup (Product, Offer, FAQ, Review),
• strictly align on‑site data with feed data,
• structure product content factually and unambiguously,
• eliminate inconsistencies between product pages, legal terms, and GMC feeds,
• ensure strong technical performance, especially on mobile,
• generate and maintain structured data directly from the feed.

Solutions like Smart Asset are designed precisely for this purpose. By leveraging enriched product feeds, Smart Asset uses best‑in‑class multimodal AI models to generate structured, consistent, and agent‑readable data. The site becomes readable, understandable, and recommendable within the future web and commerce landscape.

5. Turn the product feed into a commercial API

In an agentic world, the feed must be treated as a machine‑to‑machine commercial API, not as a simple export for media platforms. AI agents compare, rank, and recommend products based on explicit rules. The clearer, more structured, and more comparable the data, the more reliable the agent’s decisions.

What to do concretely:

• ensure each product page is understandable without visual context,
• explicitly structure shipping, return, and payment rules,
• enrich differentiating attributes (benefits, use cases, compatibilities),
• test the impact of enrichment on AI visibility and recommendation,
• treat the feed as a living product asset, not a static file.

6. Adapt performance measurement to a clickless world

Agentic commerce mechanically reduces traditional signals: fewer sessions, less navigation, fewer measurable touchpoints. Performance does not disappear—but it becomes less observable through traditional analytics tools. Product data once again becomes the anchor for analysis.

What to do concretely:

• measure performance at the product (SKU) level rather than the session level,
• reconcile Ads data, product feeds, and back‑office sales,
• identify products naturally compatible with agentic selling,
• monitor gaps between exposed products and those actually recommended.

7. Preserve the customer relationship beyond the site

When the purchase interface is operated by AI, the main risk is not traffic loss, but relationship dilution. The brand does not disappear—but it is not always visible at the critical moment of transaction. The relationship must therefore be rebuilt after purchase and outside the site.

What to do concretely:

• enrich transactional emails with brand elements,
• build loyalty programs independent of the purchase channel,
• centralize customer knowledge outside AI platforms,
• define brand tone, rules, and responses usable by agents (e.g., Business Agent).

8. Organize internal governance

Agentic commerce cannot be treated as an isolated project owned by a single team. It simultaneously impacts data, technology, compliance, and customer experience. In many organizations, product data remains fragmented across teams and tools. This fragmentation becomes a critical blocker once AI agents rely on that data to make commercial decisions. Governance therefore becomes a strategic issue, on par with infrastructure or media strategy.

What to structure first:

• align e‑commerce and merchandising on product definition,
• coordinate acquisition and data teams around product‑centric KPIs,
• involve IT, ERP, and PIM teams in data quality and synchronization,
• integrate legal and compliance teams around rules exposed to agents,
• assign clear ownership of product data,
• define continuous validation and update processes.

Product data must become a shared strategic asset, with clear and accountable ownership.

9. Test early to learn faster than others

UCP and ACP are still being shaped. Standards, usages, and consumer behaviors are not yet fixed. In this context, competitive advantage will not come from late, large‑scale adoption, but from the ability to learn faster than competitors. Brands that test early will accumulate insights that are difficult to catch up with.

Recommended approach:

• select a simple, representative category,
• activate an enriched feed on a limited scope,
• measure impact at the product level—not just globally,
• observe agent recommendation mechanisms,
• iterate quickly on attributes, rules, and content,
• progressively scale to other categories.

Testing early turns uncertainty into learning—and learning into durable advantage.

Conclusion

UCP and ACP are not incremental technical evolutions or isolated initiatives from Google and OpenAI. They signal a deep paradigm shift: the transition from click‑driven commerce to commerce driven by understanding, recommendation, and automated decision‑making.

In this new landscape, the question is no longer if agentic commerce will take hold, but how fast—and how prepared organizations will be. AI agents do not replace brands, but they radically redefine control points: interfaces matter less; data matters more.

In the medium term, three structural trends are becoming clear:

1. the standardization of product feeds as the universal commerce foundation (UCP today, ACP tomorrow),
2. the transformation of the e‑commerce site into a semantic, structured reference layer for AI agents,
3. the rise of product‑centric performance analysis, beyond traffic‑ and journey‑based metrics.

Inaction is the only real risk. Organizations waiting for perfectly stabilized standards will enter catch‑up mode, while those structuring product data today will build durable advantage.

Concretely, this means acting now on three fronts:

1. structuring and enriching product data so it is readable, comparable, and exploitable by AI agents,
2. tooling that data (feeds, site, content) so it flows frictionlessly across platforms, agents, and internal systems,
3. building a culture of testing and iteration, focused on product learning rather than channel compliance.

Agentic commerce cannot be declared—it must be prepared. It is not activated by a technical switch, but by a progressive transformation of how brands think about, govern, and exploit product data.

This guide is not a fixed roadmap, but a decision framework—a starting point for regaining control in an ecosystem where AI agents will soon become the primary demand prescribers.

The real question is no longer: are you ready for UCP or ACP?
But rather: are you ready to let agents decide for you—or to give them the right information to decide in your favor?

‍