What is agentic commerce? The UK retailer's guide

A shopper can already ask an AI assistant for something highly specific: “Find me a lightweight waterproof jacket for walking the dog that packs small and is under £120.” The assistant can interpret the request, research or retrieve products, compare options and give a short recommendation.

That request is the kind SO-ME, the consumer shopping platform Sighted powers, receives: a use, a constraint and a budget rather than a category name. Matching a request put that way to the product information retailers already hold is difficult, which is why the product data section below carries more weight than the protocol section.

On some platforms, for some participating merchants, the journey goes further into native or agent-enabled checkout. Maturity varies by platform and country. Discovery and recommendation are already widespread. Native checkout is real but not universal.

When a shopper browses a category page, they interpret images, filters, descriptions and brand cues themselves. When an AI system mediates the decision, the retailer needs enough accessible product information for that system to tell whether the product meets the shopper's constraints. That makes product information infrastructure strategically important:

• catalogue accuracy;
• attributes;
• variants;
• price and stock freshness;
• product-page facts;
• Merchant Center data;
• commerce feeds;
• compatibility and use-case information;
• delivery and fulfilment signals where relevant.

This does not replace SEO or brand. It adds a layer of machine-mediated product selection.

One Sighted result, in its own context and not a universal benchmark: in the pilot with The Game Collection, 200 products were enriched and 17 collections created; AI citation coverage went from 0% to 32% of the priority products tested, and the store's organic conversion rate rose 64%.

Adobe reported in April 2026 that traffic from AI sources to US retail sites grew 393% year on year in the first three months of 2026, and that in March 2026 that traffic converted 42% better than non-AI traffic. Those figures are US retail data, not a global benchmark, but they show why the channel deserves attention.

Salesforce reported that AI and agents influenced 20% of retail sales over the 2025 holiday season (1 November to 31 December), $262 billion, through mechanisms including personalised recommendations and deeper customer engagement. “Influenced” is the important word: it is not a claim that autonomous agents placed one in five orders.

McKinsey projects (October 2025) that AI agents could orchestrate $3 trillion to $5 trillion of global consumer goods revenue, and $900 billion to $1 trillion of US retail, by 2030.

The UK is already an active market for AI-led discovery, while some native checkout programmes have yet to reach it. Google's Merchant Center help says the checkout UCP enables is available to select merchants in the US, Canada and Australia, and Google has said the UK comes later, without a date.

OpenAI supports product discovery through ChatGPT and ACP-based merchant data today, while checkout availability depends on the merchant and the market.

That gives UK retailers a useful window: the discovery behaviour is already here, while the transaction layer is still developing. The sensible response is to make your product information accurate, useful, portable and measurable now, rather than building speculative infrastructure for every protocol.

Google's guidance on its AI features says established SEO fundamentals remain relevant: good crawlability, helpful content in text, internal linking, images and video, structured data that matches the page, and accurate Merchant Center information. The added discipline is making products usable inside decision systems, as well as making pages visible in search results.

ACP, the Agentic Commerce Protocol

OpenAI's open commerce protocol, co-developed with Stripe, used for merchant and product discovery and commerce integrations in ChatGPT.

UCP, the Universal Commerce Protocol

A Google-backed open standard for commerce actions across AI Mode, Gemini and related surfaces.

MCP, the Model Context Protocol

A broader standard, introduced by Anthropic, for connecting AI applications to tools and data. It matters for agent infrastructure and does not replace a product feed.

A2A, AP2 and other standards

Adjacent agent and payment protocols that may matter depending on the platform and the integration.

Is agentic checkout live?

Yes, on selected platforms, for selected merchants and in selected markets. It is not yet universal. Discovery and recommendation are far more broadly established than native checkout.

Do I need to implement ACP and UCP immediately?

Not necessarily. Start with your existing catalogue, website and merchant feeds. Protocol implementation depends on market, platform, merchant eligibility and technical strategy.

Is agentic commerce the same as AI shopping?

AI shopping is the consumer behaviour and experience. Agentic commerce is the wider ecosystem of discovery, decision, commerce actions, payment and fulfilment mediated by AI agents and commerce systems.

1. Fix product-data fundamentals. Agentic commerce does not excuse poor catalogue hygiene: price, stock, identifiers, variants, taxonomy and product facts still need to be right.
2. Understand conversational demand. Analyse how shoppers describe needs, occasions, constraints and trade-offs that your existing product data does not capture well.
3. Publish the right information to the right surfaces: product-page content, structured product data, Merchant Center attributes, platform and product feeds, or commerce integrations.
4. Measure product selection, not brand mentions alone: whether the right products are surfaced and recommended for the prompts they should win. For ChatGPT, the ranking factors guide lists what OpenAI documents about that selection.
5. Connect selection to commercial outcomes. Track traffic, conversion and revenue for treated products against a baseline and, where possible, comparable untreated products.

Sighted sits between your catalogue and the AI shopping ecosystem. It benchmarks where products appear, diagnoses why relevant products are being missed, enriches the underlying data at scale, distributes approved changes into your existing channels and measures whether recommendation and revenue improve. The free AI visibility audit is the benchmark step on up to 50 of your product pages; on Shopify the approved values are written back to the store. The enrichment is the technology behind SO-ME, which meets the same requests from the shopper's side.