AI in eCommerce: How Intelli gent Commerce Is Rewriting the Customer Journey

AI-referred traffic to U.S. retail sites grew 4,700% year-over-year in 2025. Orders from AI-powered searches on Shopify alone increased 15x. The global AI in eCommerce market is on its way to $64 billion by 2034. This is not a trend to monitor, it is a structural shift happening right now, and the brands moving early are already claiming the shelf space that late movers will find taken.

The way people shop online is undergoing the most significant shift since mobile commerce. Artificial intelligence is no longer a feature bolted onto an existing experience, it is becoming the foundation of how customers discover, evaluate, and purchase products. And the numbers have reached the point where ignoring them is no longer a defensible position for any eCommerce brand.

AI-referred traffic to U.S. retail sites grew 4,700% year over year in 2025. Orders from AI-powered searches on Shopify alone increased 15x in the same period. The global AI in eCommerce market, valued at $8.65 billion in 2025, is projected to reach $64 billion by 2034. Revenue per visit from AI-referred shoppers jumped 254% year over year during the 2025 holiday season alone. These are not indicators of a technology trend, they are the fingerprints of a fundamental restructuring of how commerce happens.

For brands still treating AI as an optional add-on or a marketing experiment, the window to catch up is narrowing in real time. The brands moving deliberately today are not just improving metrics on their existing storefront, they are building presence in an entirely new discovery and purchase layer that did not exist two years ago. This article explains exactly what that layer is, how it works, and what brands need to do to claim their position in it.

For decades, the customer journey had a predictable starting point: a search bar. Whether it was Google, Amazon, or a brand’s own site search, the pattern was consistent, a shopper typed a query, received a list of links, clicked through, compared, and eventually bought. That model is being replaced, and the replacement is moving faster than most eCommerce teams have absorbed.

Shoppers are increasingly turning to AI assistants, ChatGPT, Google’s Gemini, Perplexity, and a growing ecosystem of specialized agents, to answer shopping questions that would previously have started a multi-day search journey. Questions like “What’s the best running shoe for flat feet and heavy pronation under $120?” or “Find me a non-toxic mattress that ships to Canada with good reviews for side sleepers.” These assistants do not return a list of ten blue links. They return a curated answer, with product recommendations, with comparative reasoning, and increasingly with a direct path to purchase, all within the same conversation.

This is agentic commerce: AI acting as a co-pilot, and sometimes a complete proxy, for the shopper. The shopper asks a natural language question. The agent reasons about it, queries available product data, cross-references reviews and inventory, applies personalization signals, and presents recommendations with one-tap checkout. The brand’s website may never appear in this journey at all.

“Your product data, descriptions, dimensions, ingredients, pricing, availability, is now your shelf space in an AI-powered store. If the agent cannot read your data, it cannot recommend your product.”

— AI in eCommerce: Intelligent Commerce Framework, 2025

The clearest early signal of how far this has already moved was OpenAI’s Instant Checkout, launched in September 2025. ChatGPT can now recommend a product and execute the transaction on the user’s behalf, without the customer ever visiting the brand’s website. A shopper can ask ChatGPT to find and buy a gift, approve the recommendation, and complete the purchase in a single conversational thread. Similarly, Google’s AI Mode in Search can now surface Shopify merchants’ products and enable checkout directly inside the Gemini app, bypassing the traditional click-to-website pattern entirely.

The implication for brands is not subtle. Investing heavily in on-site conversion rate optimization for a discovery channel that a growing share of high-intent shoppers are bypassing entirely is a strategy with diminishing returns. The investment priority is shifting: structured product data, AI-ready feeds, and protocol compatibility are becoming more valuable than homepage design. The storefront is still important, but it is no longer the only arena that matters.

Behind every consumer-facing AI commerce experience, the Gemini recommendation, the ChatGPT checkout, the Perplexity product comparison, is a new piece of shared infrastructure that most brands have not yet heard of but will need to understand urgently. It is called the Universal Commerce Protocol (UCP), and it is the foundational standard that makes agent-driven commerce interoperable at scale.

UCP was co-developed by Google and Shopify, and has since been endorsed by a coalition of commerce, payments, and retail leaders that signals just how seriously the industry has committed to this direction.

Think of UCP as the HTTP of AI commerce. Just as HTTP gave web browsers and servers a shared language to communicate, enabling any browser to render any website without a custom integration between them, UCP gives AI agents a shared language to discover merchant capabilities, negotiate what they can handle, and complete transactions without each agent needing a custom integration with every retailer.

Without UCP, an AI agent wanting to sell products from a specific retailer would need a bespoke API integration, a custom authentication flow, a proprietary data format mapping, and a retailer-specific checkout implementation. Multiply that by thousands of merchants and dozens of AI platforms, and the resulting fragmentation would make agent commerce practically impossible at scale. UCP solves this by standardising the protocol layer, so any agent built on the standard can interact with any merchant on the standard, automatically.

“UCP is the HTTP of AI commerce, the shared protocol that makes the entire ecosystem interoperable. The brands that optimise for it early will have the equivalent advantage of being discoverable on the early web.”

— Universal Commerce Protocol, co-developed by Google and Shopify, 2025

The compression of the shopping funnel that AI enables is not a marginal improvement in an existing journey, it is a qualitative restructuring of how purchase decisions get made. The before-and-after contrast is stark enough that it is worth walking through in detail, because the operational implications for brands are significant.

The conversion rate difference is the metric that matters most for brands trying to make the business case for AI investment. AI-engaged shoppers already convert at 12.3% versus 3.1% for non-AI shoppers, a near 4x lift. This is not because AI agents are better salespeople. It is because they collapse the consideration phase that accounts for the vast majority of traditional funnel drop-off. By the time a shopper confirms a purchase in an AI conversation, they have already received the comparison, the social proof, and the compatibility check that previously required multiple sessions across multiple sites.

The implications go well beyond conversion rate. Brands that invest in clean, structured product data and AI-ready architecture will appear in these agentic results. Brands that do not, with incomplete attribute data, missing schema markup, unstandardised feeds, or no UCP compatibility path, will effectively become invisible in the fastest-growing discovery channel in eCommerce. Not less visible. Invisible.

The visibility cliff
An AI agent querying UCP-connected merchants for a specific product returns results based on the quality and completeness of structured product data. A merchant with incomplete attributes, missing dimensions, or unstructured descriptions simply does not appear in the results, not because they were ranked lower, but because the agent could not parse their data well enough to include them in the recommendation. This is a binary outcome: you are in the results or you are not. There is no position 8.

Agentic commerce, AI agents completing transactions on behalf of shoppers, is the most visible and dramatic manifestation of AI in eCommerce. But it is not the only one generating measurable revenue impact right now. AI personalization is already a significant and growing revenue driver for brands that have invested in the infrastructure to support it, and the ROI data is unusually clear for an emerging technology.

AI personalization already drives 25–35% of eCommerce revenue for leading retailers. 89% of brands implementing AI personalization report positive ROI with an average payback period of just nine months. These are not aspirational projections, they are reported outcomes from brands operating production personalization systems today.

Real-time recommendation engines that factor in browse history, purchase patterns, session context, and similarity signals, surfacing products each individual shopper is most likely to convert on.

Drives 35% of Amazon revenue

ML models that adjust pricing in real time based on demand signals, competitor pricing, inventory levels, and customer segment, maximising margin without manual intervention.

Up to 25% margin improvement

Predictive models that reduce overstock and stockout events by forecasting demand at the SKU level, factoring in seasonality, trend signals, and promotional calendars.

20–30% inventory cost reduction

Agentic customer service systems that handle order enquiries, return initiations, size recommendations, and product questions, resolving 60–80% of support volume without human intervention.

60–80% ticket deflection rate

Automated generation of product descriptions, meta titles, alt text, and structured attributes, keeping product data complete, consistent, and AI-agent-readable at catalog scale.

Catalog readiness for AI agents

Real-time transaction scoring that identifies fraudulent patterns without blocking legitimate purchases, reducing both fraud losses and false positive friction that drives conversion drop-off.

40–60% fraud reduction reported

What connects all of these applications is the same underlying requirement: clean, structured, well-governed data. Recommendation engines require accurate purchase and browse history. Dynamic pricing requires reliable competitor feeds and inventory signals. Demand forecasting requires consistent historical sales data at the SKU level. The brands generating 25–35% of revenue from AI personalization did not get there by deploying sophisticated models on messy data. They got there by treating data quality as a prerequisite to AI investment — not an afterthought.

The shift from search-driven commerce to agent-driven commerce is not a future scenario to prepare for. It is happening now, built on open standards like UCP and powered by platforms like Shopify, Google, and OpenAI. The question is not whether to act, it is how to sequence the investment to capture the most value in the shortest time. Here is the four-part action plan, in priority order.

Schema markup, consistent attribute naming, accurate inventory feeds, complete dimension data, structured ingredients or material specifications, this is the foundation for appearing in AI-driven results. An AI agent querying UCP-connected merchants returns recommendations based entirely on the quality and completeness of your structured product data. If an agent cannot read your data, it cannot recommend your product. The audit should answer three questions: Is every product attribute complete and consistently formatted? Is the data accessible via a machine-readable feed? Does it update in real time when inventory or pricing changes?

Start here: Run your product catalog through Google’s Rich Results Test and Structured Data Markup Helper to identify schema gaps before they translate into agentic invisibility.

Shopify merchants are already inside the UCP ecosystem, the protocol compatibility is built into the platform, and product catalogs are automatically accessible to any UCP-compatible agent without additional development work. If you are on Adobe Commerce, BigCommerce, Salesforce Commerce Cloud, or a custom stack, evaluate your path to UCP compatibility explicitly. At minimum, ensure your product feeds are structured, machine-readable, and updated in real time. The merchants who delay this evaluation are the ones who will be asking why their AI traffic is flat while competitors are seeing 15x order growth from the same channel.

Shopify merchants: verify your Storefront API and product feed are fully configured, UCP compatibility requires complete attribute data, not just catalog access.

AI personalization already drives 25–35% of eCommerce revenue for leading retailers, with 89% of implementing brands reporting positive ROI and a nine-month average payback period. The investment compounds: the data you collect from early personalization deployments becomes the training signal that improves recommendation quality over time. Every month you delay personalization investment is a month of customer behavior data you are not collecting. The starting point does not need to be sophisticated, a well-implemented product recommendation engine on collection and product pages, fed by clean behavioral data, generates measurable lift within the first quarter.

The 9-month average personalization ROI payback means an investment made today returns capital before the end of the year, while building the data foundation for increasingly sophisticated applications in subsequent years.

The next era of eCommerce is definitionally channel-agnostic. Products will be discovered and purchased through AI assistants, voice interfaces, social agents, smart devices, and AR environments. Brands that have built their commerce experience exclusively around a website, with brand expression, persuasion architecture, and conversion optimization all assuming the customer is on their domain, are building for a single channel in a multi-channel world. Your brand experience, your product data, and your checkout capability need to work across all of them. This does not mean abandoning your website. It means ensuring that the data and commerce infrastructure beneath it is portable and API-first enough to power every surface where your customers might encounter your products.

The brands that will own AI commerce shelf space are those treating their product catalog as infrastructure, not content. Clean, structured, real-time, and API-accessible is the new competitive moat.

The shift from search-driven commerce to agent-driven commerce is already underway, built on open standards like UCP and powered by platforms like Shopify, Google, and OpenAI. The conversion data, 12.3% versus 3.1% for AI-engaged versus non-AI shoppers, is not a projection. The 4,700% YoY growth in AI-referred traffic is not a forecast. The 15x increase in Shopify orders from AI-powered searches is not an estimate. These are production numbers from 2025, and they reflect a commerce environment that has structurally changed.

Brands that move early will own the shelf space in this new layer. They will earn agent visibility through data quality, build personalization systems that compound in value as behavioral data accumulates, and create brand experiences that work across every surface where their customers shop, not just the one they designed in 2019. The ones that wait will find that shelf already taken, and the path to reclaiming it will be longer and more expensive than the investment required to claim it today.

The product data audit. The UCP compatibility check. The personalization infrastructure investment. The channel-agnostic commerce architecture. None of these are complex initiatives. All of them have shorter time-to-value than the traditional eCommerce investments that will deliver diminishing returns as agentic commerce grows. The question is not whether AI is changing eCommerce. It already has. The question is whether your brand will be visible when the agent comes looking.

AI is transforming the customer journey by replacing traditional search-based shopping with conversational and personalized experiences. Customers can now discover products through AI assistants, receive tailored recommendations, compare options instantly, and even complete purchases without visiting multiple websites.

Agentic commerce is AI acting as a complete proxy for the shopper, not just surfacing recommendations within a brand’s existing experience, but independently querying products, comparing options, and completing transactions on the shopper’s behalf. Regular AI recommendations (like Amazon’s “Customers also bought”) exist within a brand-controlled surface where the shopper is already engaged. Agentic commerce happens entirely outside the brand’s website, within ChatGPT, Gemini, or a third-party agent, meaning the brand may never see the shopper visit their site at all. The purchase happens through the agent’s interface, powered by the brand’s structured product data and UCP compatibility.

For a Shopify brand with a reasonably well-maintained product catalog, achieving a baseline AI-ready state, complete attribute data, schema markup, accurate feeds, UCP ecosystem access, is a 4–8 week project, not a multi-quarter initiative. The timeline extends for brands with large catalogs, inconsistent historical data entry practices, or platforms requiring custom UCP integration work. The bottleneck is almost always data quality rather than technical complexity: the technical integrations are relatively straightforward; the work of cleaning, completing, and standardising product data at catalog scale is where time is spent. Starting with your highest-traffic, highest-revenue product categories and expanding from there is the fastest path to measurable AI commerce lift.

The UCP ecosystem and agentic commerce are explicitly designed to be accessible to brands of all sizes, the protocol standardisation that makes agentic commerce work is, by design, non-discriminatory toward brand size. A mid-market Shopify merchant with complete, well-structured product data competes on equal terms with a large retailer in an AI agent’s recommendation set. The competitive variable in agentic commerce is data quality, not marketing budget. This is one of the most significant structural differences from traditional paid search, where larger budgets produce higher visibility. In agent-driven discovery, visibility is earned through data completeness, which is a capability investment accessible to brands at every scale.