Agentic commerce vs traditional ecommerce: what actually changes

What is the difference between agentic commerce and traditional ecommerce?

In traditional ecommerce, the shopper does every step personally: search, browse, compare, read reviews, check out, all on interfaces designed for human eyes. In agentic commerce, an AI shopping agent helps with those steps: it researches, compares, answers questions, and narrows the options, and on some platforms can complete the checkout once the shopper decides.

The core shift is who does the legwork, not who decides or where the sale happens. In both models the merchant usually remains the seller of record, and in both the shopper stays in charge of the choice. And an agent is not a chatbot: chatbots advise, agents execute a chain of tasks end to end, from query to a cart the shopper approves.

The forecasts are serious. McKinsey projects agentic commerce could reach $3 trillion to $5 trillion in orchestrated global retail revenue by 2030. Even the bottom of that range justifies preparing now.

How does each stage of the purchase change?

Six stages of the purchase change. The table below is the whole argument in one place.

Checkout moved most in the past year. On September 29, 2025, Stripe and OpenAI launched Instant Checkout and the open Agentic Commerce Protocol, with Etsy live at launch. Under ACP the merchant still accepts or declines each order, processes payment with its existing provider, and owns fulfillment and support. The protocol changes where the buy button appears, not who runs the store.

Loyalty changes more quietly. A human reorders out of habit; an agent re-runs the comparison every time. Habit gives way to the earned default: clean data, reliable stock, honest pricing.

Which metrics change when AI agents shop?

Traffic quality replaces traffic volume as the headline metric. Fewer sessions arrive, but they arrive decided: Adobe Analytics measured AI-referred traffic to US retail sites up 138% year over year in May 2026, and they engage more deeply because qualification happened before the click.

The conversion story deserves the honest version, since most coverage cites one snapshot. In July 2024, generative-AI traffic to US retail sites converted 43% worse than other channels. By February 2025 the gap was 9%. By June 2026 it had flipped decisively positive. That is a channel maturing, not a fad.

Attribution breaks at the same time: agent research happens off your site, so last-click models undercount AI influence and AI-referred visits land as dark traffic unless you instrument referrers. Four KPIs worth adding now:

• AI-referral share of sessions and revenue
• Revenue per AI-referred visit
• Citation share: how often ChatGPT, Perplexity, and Gemini recommend you in your category
• Feed coverage: the share of your catalog an agent can parse

What stays the same?

More than the hype suggests. Under the Agentic Commerce Protocol the plumbing is unchanged: the merchant accepts or declines each order, charges through its existing payment provider, ships, and handles returns and support. Losing the customer relationship is the most overstated fear.

The fundamentals matter more, not less. An agent cannot be charmed by a hero image or a discount popup. It compares price, stock reliability, shipping speed, and return policies on every purchase. Strong operations make agents your best channel; weak ones get routed around.

Margins and unit economics do not change because the buyer is a machine. And brand still matters upstream: consumers tell agents which brands to consider. The brand you built for people is the agent's shortlist.

Do agentic and traditional ecommerce coexist?

Yes, and the mix is the point. Salesforce measured AI and agents influencing $67 billion in purchases during Cyber Week 2025, 20% of all global online orders. Read the other way: 80% were still conventional. This is a shifting mix, not a replacement event. People still like to browse, especially for considered purchases: a sofa, a pair of glasses, a jacket you will wear for years. Agents compress the research, not the desire to choose.

The market corrected once already. In March 2026, OpenAI moved Instant Checkout out of inline ChatGPT responses into merchant-branded apps, after merchants flagged the transaction complexity that lives on their side: subscriptions, inventory, shipping, taxes. The dominant 2026 pattern: discover and decide in AI, complete on the merchant's site.

That makes your own domain the convergence point: one storefront must serve human browsers and stay legible to agents. With ACP, Mastercard Agent Pay, and Visa Intelligent Commerce all in market, multi-protocol readiness is the baseline.

When does each model win?

Delegation is not evenly distributed across what people buy. Treat this as a portfolio decision.

Where agentic commerce wins today

• Replenishment and reorders: the agent already knows the winner
• Spec-driven comparison shopping, where the decision is a table, not a feeling
• Price and deal hunting across many stores at once
• Time-poor, high-intent buyers who want the errand done
• Research-heavy purchases, which Adobe found shoppers far more likely to hand to AI

Where traditional ecommerce still wins

• Taste- and identity-driven categories, where browsing is the pleasure
• First-time brand discovery, where the story sells the product
• Try-before-you-buy and fit-sensitive purchases
• Impulse and inspiration shopping

Score your catalog by delegability: agent-readiness where delegation is high, human conversion work where it is low.

How do you prepare your store for agentic commerce?

Preparation means making the store machine-answerable. Agents reward what comparison engines always have, only more literally. An agent-readable storefront on your own domain keeps you the seller of record while making every product legible to the agents doing the research.

• Expose structured product data: prices, availability, variants, specs, policies
• Keep product feeds complete and accurate; feed coverage is the new crawl budget
• Publish clear shipping, returns, and warranty policies on stable URLs
• Keep pricing and availability consistent across pages, feeds, and protocol endpoints
• Track AI referrers separately so agentic traffic stops hiding in direct
• Ask ChatGPT and Perplexity to compare products in your category; note whether you are cited or skipped

Measure before you build: find out where AI engines already send your category's traffic and whether agents can read your product pages today. Every row of the table above is a test you can run this week.