AI Shopping: From Discovery to Purchase with UCP and JD Sports

The era when search engines merely influenced shopping is giving way to one where AI agents can complete purchases on a shopper’s behalf — and recent moves by JD Sports, Google, Shopify, Microsoft and others make that shift unavoidably practical rather than hypothetical. JD’s decision to let U.S. customers search for and buy products directly through AI platforms, combined with Google and Shopify’s launch of the Universal Commerce Protocol (UCP), crystallises what many retailers have suspected for years: AI shopping is moving from discovery to transaction, and the winners will be those who strengthen fundamentals while experimenting strategically.

Overview: what changed, and why it matters​

AI shopping — often called conversational commerce or agentic commerce — is the model where a conversational AI (an “agent”) performs discovery, recommends options, and can be authorised to complete payment and fulfillment steps on behalf of a user. The practical turning point in early 2026 was twofold: first, major platforms and vendors aligned on standards and product integrations that make in-chat checkout operational at scale; second, a big omnichannel retailer publicly committed to connecting its catalog directly into those agentic channels. Together those developments move AI from a research-and-marketing conversation to an operational imperative.
Google’s Universal Commerce Protocol (UCP) attempts to define a common language for agents, merchants, payment providers and fulfillment systems so that an AI on one device can reliably access product data, check availability, and trigger secure payments across vendor boundaries. Shopify has positioned itself to connect merchants to AI conversations centrally, and Microsoft, OpenAI and others are rapidly adding in-chat purchase flows. That means the old separation — search results on one side, checkout on another — is dissolving into a conversational pipeline that can be end-to-end.
JD Sports’ U.S. initiative — built on a commercetools replatforming and payment integrations via Stripe — is the clearest example to date of a major retailer treating AI as a transactional channel rather than a mere referral source. JD intends to let shoppers move from product discovery to one-click purchase inside Copilot, Gemini and ChatGPT-like experiences. That step is an operational proof that agentic commerce is not only technically feasible but commercially attractive for brands with investment-ready digital stacks.

---

Background: agentic commerce, standards and commercial incentives​

What is agentic commerce?​

Agentic commerce describes experiences where AI agents act autonomously or semi-autonomously to carry out shopping tasks. These agents can be:

• Personal assistants (on-device or cloud-powered) that manage lists, preferences and payment authorisations.
• Platform agents (Copilot, Gemini, ChatGPT) that aggregate product feeds and present recommendations.
• Merchant-hosted agents integrated directly into brand systems.

The common denominator is that agents bridge discovery, selection and payment into a conversational, often single-interaction flow. That creates both convenience and concentration of power: agents determine which products are surfaced and which merchants get the buy button.

Why standards like UCP matter​

A fragmented agent ecosystem would require merchants to build bespoke integrations for each major AI — an impossible cost curve for most retailers. UCP aims to reduce that friction by creating:

• A common data model for product metadata, pricing and availability.
• Defined payment and fulfillment primitives that agents can call.
• Compatibility with existing retail infrastructure (APIs, inventory systems, payment gateways).

If broadly adopted, UCP lowers the integration barrier and makes agentic channels more accessible to mid-market and enterprise retailers alike. But the specification also creates a leverage point: the organisation that controls the dominant protocol or its reference implementation can shape incentives and capture economic rents.

---

SEO and discovery: fundamentals don’t vanish — they evolve​

“SEO is not dead” — but it must be reinterpreted for conversational layers​

Every major change in search has prompted premature obituaries for SEO. The introduction of agentic shopping changes how signals are consumed, not which signals matter. Under the conversational layer, discovery still relies on:

• Structured product data (consistent titles, attributes, SKUs).
• Authority and trust signals (reviews, brand reputation).
• Relevance signals (category taxonomy, internal linking, historical conversion rates).

AI agents abstract keywords and replace them with natural-language intents, but retrieval uses the same plumbing — structured feeds, product APIs, and performance metrics. The risk is that brands assume AI will “find” them without the discipline of consistent data governance; in practice, poorly structured catalogs disappear inside agentic recommendations.

What changes for search engine optimisation practitioners​

• Product titles must be precise and consistent across channels; synonyms and marketing flourishes belong in supporting copy, not canonical feeds.
• Schema markup and machine-readable attributes become first-class outputs of merchandising teams.
• Internal linking and taxonomy retain importance because they provide entity context that agents use to understand relationships (complements, substitutions, compatibility).
• Historical performance and trust metrics will be surface signals for agents deciding which brands to recommend to a user.

In short: SEO’s strategic objective — to be discoverable for relevant intent — remains. The tactics shift toward feed quality, structured metadata and cross-channel alignment.

---

Ecommerce performance: the revenue engine that still pays the bills​

Don’t let agentic novelty distract from conversion fundamentals​

Agentic channels magnify friction. If an AI surfaces inaccurate stock levels, slow checkout pages, or wrong pricing, the result is lost trust and immediate revenue damage. For many UK retailers, the urgent task is not to ship to every AI today but to fix:

• Site speed and mobile performance.
• Internal search quality and faceting.
• Consistent pricing and availability across channels.
• Accurate, governed product information.

JD’s move succeeds because it rests on a recent and material replatforming that strengthened its product APIs and checkout flows — investments many retailers have yet to make. Agentic commerce can amplify competitive advantages only when core ecommerce systems already work reliably.

Analytics and attribution in an agentic world​

Standard analytics models (last-click, session-based funnels) were never perfect for multi-touch journeys; agentic shopping makes them more brittle. Retailers must:

• Instrument product feeds and API calls so agent-origin traffic is taggable and traceable.
• Track intent-to-conversion at an event level, logging the full conversational context that led to a purchase.
• Use first-party identity where possible to stitch agent interactions to customer profiles without over-relying on third-party cookies.

Without granular observability, AI-driven behaviour will be guessed at, not measured — making optimisation and ROI decisions speculative.

---

The fragmentation question: one standard or many?​

Expect coexistence, not instant consolidation​

There is a temptation to treat UCP as the future winner. It’s important to be realistic: platform economics, vendor incentives and legacy systems push toward fragmentation. Shopify and Google push UCP because it accelerates agentic commerce adoption in ways that benefit their merchants and ad ecosystems. Microsoft and other large vendors are simultaneously building proprietary integrations and embedded checkout experiences. Commercetools, Stripe and other commerce platforms are offering retailer-controlled alternatives. The practical result is likely years of coexistence — platform-specific capabilities, vendor-specific optimisations, and hybrid models where merchants expose both UCP-compatible endpoints and bespoke integrations.

What this means for retailers​

• Flexibility beats allegiance. Don’t pin your roadmap on a single protocol; design modular APIs and exportable product feeds to adapt as agent preferences shift.
• Data-first architecture. Prioritise canonical product data and a single source of truth that can feed UCP, platform APIs and bespoke connectors.
• Strategic vendor partnerships. Use platform partnerships (like commercetools or Shopify) to accelerate integrations, but keep strategic control over pricing, loyalty and customer ownership.

---

Case study: JD Sports — strengths, risks and lessons​

Why JD’s approach is instructive​

JD’s U.S. initiative shows a textbook implementation of agentic commerce:

• A recent replatforming onto a modern commerce stack (commercetools) that exposes strong product APIs.
• Direct payment integrations (Stripe) to enable secure one-click purchases.
• Collaboration with major AI platforms to enable discovery and purchase without redirecting users away from agents.

This architecture reduces friction for customers and centralises the merchant’s control over product, pricing and fulfillment logic — which is essential when your catalog is surfaced by an external agent.

Risks that JD and any early adopter face​

• Regulatory and consumer-protection exposure. Inaccurate availability claims or pricing errors in agentic purchases can trigger complaints and regulatory scrutiny across jurisdictions with stronger consumer protections.
• Margin pressure and fee leakage. Embedded agent checkouts may carry variable fees or require promotions to get preferential placement, compressing margins.
• Data dependence and platform leverage. The better agents perform, the more merchants may rely on platform-level analytics and incentives, raising the risk of losing direct first-party relationships with customers.

Lessons for UK retailers (or any cautious adopters)​

• Prioritise data governance and product feed hygiene before enabling agentic channels.
• Pilot agentic experiences in controlled markets or product lines where inventory logic and fulfillment are simplest.
• Maintain ownership of checkout and loyalty mechanics to protect lifetime value.

---

Practical checklist: preparing for AI shopping without sacrificing fundamentals​

Below is a prioritized set of actions that ecommerce teams can execute with moderate effort and high impact.

• Product data and feeds
• Standardise product titles and SKUs across channels.
• Surface variant attributes (size, colour, material) as discrete fields in feeds.
• Implement schema.org markup and ensure real-time availability endpoints.
• Platform and site performance
• Audit and improve time-to-interactive and mobile rendering metrics.
• Harden internal search relevance and faceting logic; surface intent-friendly synonyms.
• Validate pricing logic across country, channel and loyalty tiers to avoid agent-sourced mismatches.
• Analytics and observability
• Tag agent-origin sessions and capture conversational context events.
• Instrument server-side APIs to log UCP/API calls, payloads and outcomes.
• Map agent interactions into your existing customer data platform (CDP) or CRM.
• Architecture and integrations
• Adopt a modular, API-first commerce layer that can expose multiple endpoints (UCP, platform APIs, direct integrations).
• Keep payment and loyalty control in-house where possible (tokenised payments, SSO).
• Use middleware to translate and mediate between internal models and external protocols.
• Strategy and governance
• Run small proofs of value before committing to broad agentic rollouts.
• Align legal and privacy teams early to define consent, data retention and refund processes.
• Set clear KPIs for agentic investments (incremental revenue, LTV impact, cost-per-acquisition).

---

The regulatory, privacy and trust calculus​

AI shopping raises specific trust vectors. Agents can obscure provenance (which merchant is actually selling the product), compress or alter the presentation of terms, and automate transactions in ways consumers may not expect. That leads to several imperative considerations for retailers:

• Transparent presentation. Ensure agents can surface your branding, terms, returns and warranty information at point-of-decision to avoid disputes.
• Accurate stock and pricing. Real-time inventory and price accuracy are non-negotiable; even occasional mismatches will erode trust faster than in a traditional web flow.
• Consent and data minimisation. Work with legal to define what conversational context you may store and reuse, and disclose that clearly to customers.
• Dispute and chargeback handling. Agentic flows must map to clear, auditable records for fulfilment and refunds.

Until regulatory guidance catches up with innovation, brands that over-communicate and over-document will fare better than those that treat agentic channels as black boxes.

---

Strategic posture: experiment strategically, protect what pays​

Boards and stakeholders will push for visible AI projects. That pressure is real and sensible — but rapid, unfocused adoption can be damaging. The recommended strategic posture is:

• Defend the core. Fix product data, performance and analytics first. These deliver revenue today and are prerequisites for agentic scale.
• Run controlled experiments. Pilot one-click agentic checkout for a subset of SKUs, regions or loyalty customers to gather operational learning without jeopardising brand trust.
• Invest in modularity. Choose commerce platforms and middleware that let you expose multiple integration endpoints without replatforming.
• Focus on customer understanding. The best agentic experiences are built on first-party customer signals and clear merchandising priorities.
• Negotiate commercial terms. When integrating with platform agents, secure commercial clarity on fees, placement, data access and measurement.

This mix of defence and selective offence preserves short-term profitability while positioning the business to scale agentic commerce if and when market adoption and economics justify it.

---

Future scenarios: what to expect over the next 2–5 years​

• Scenario A — Coexistence and slow consolidation: multiple protocols and platform-specific features coexist for several years; merchants who invest in modular APIs win stable visibility.
• Scenario B — Platform consolidation: dominant platforms push feature parity into a single de facto standard (e.g., a UCP-led ecosystem); merchants who adopted earlier enjoy a “first-mover” advantage.
• Scenario C — Regulatory reset: consumer-protection and competition regulators impose constraints on agentic commerce that reshape trust and disclosure requirements, favouring brands with transparent processes.

Across these scenarios the common thread is that merchant control over product data and customer relationships will determine long-term profitability. Technology shapes distribution; it does not create demand out of thin air.

---

Final analysis: where to focus now​

AI shopping is not a technology novelty to be chased for its own sake. The real change is structural: discovery and purchase are collapsing into unified conversational flows, and protocols like UCP make that path easier for merchants with clean data and agile architecture. The immediate winners will not necessarily be the loudest first movers; they will be the retailers who:

• Maintain rigorous product data hygiene and canonical feeds.
• Fix ecommerce performance and conversion bottlenecks that agentic channels will expose.
• Keep strategic ownership of checkout, loyalty and customer data while experimenting with agentic integrations.

In short, prepare for AI shopping by doing what always pays off: invest in accuracy, speed, trust and measurement. Do that, and you’ll be ready to let an AI recommend your product — and collect the payment — without sacrificing the fundamentals that keep the lights on.
Conclusion: the arrival of meaningful agentic commerce is real, but it is not a binary threat to SEO or ecommerce fundamentals. It is an evolution that rewards the disciplined — the merchants who clean their data, own their checkout, instrument their analytics and design modular systems. Those organizations will gain visibility inside tomorrow’s AI conversations; those that chase novelty while neglecting the basics risk being unheard even when the conversation turns to them.

---

Source: InternetRetailing OPINION: Preparing for AI shopping without sacrificing the fundamentals - InternetRetailing