Agentic commerce means AI agents doing the shopping: understanding what a customer wants, researching and comparing options across the web, and — increasingly — completing the checkout, all on the shopper's behalf. It's the shift from customers visiting your store to software evaluating your store for them. For owners and marketers, this changes the three things your growth depends on: how products get discovered, what "converting" means when the visitor is an agent, and which data decides whether you're recommended at all.
What's actually happening (not the sci-fi version)
Three layers of this are already live:
1. Agents answer shopping questions — and send real traffic. Product research has moved into ChatGPT, Perplexity, Google's AI results, and Amazon's Rufus, and the numbers are no longer small: Adobe Analytics measured AI-sourced traffic to US retail sites up 693% year-over-year over the 2025 holiday season, and still growing 393% year-over-year in Q1 2026. Shoppers ask "best budget mechanical keyboard for a small desk" and get a synthesized shortlist — brands included or excluded before your site ever gets a visit.
2. Agents connect to catalogs. AI platforms and e-commerce platforms are wiring merchant catalogs directly into assistants, so agents pull live products, prices, and availability rather than guessing from old crawls. With OpenAI's Instant Checkout, over a million Shopify merchants' catalogs are being connected to ChatGPT — if you're on Shopify, your catalog is becoming agent-readable infrastructure whether you planned it or not.
3. Agents complete purchases. Agent checkout is live, not theoretical: US ChatGPT users can already buy directly from US Etsy sellers in chat, powered by the Agentic Commerce Protocol that OpenAI co-developed with Stripe and open-sourced. The merchant stays the merchant of record — processing payment, fulfilling, handling support — and pays a small fee on completed purchases, while product results are not influenced by the fee.
Each layer removes a step where your website, your merchandising, and your ads used to do the persuading.
The data says these visitors are better, not worse
The early fear was that AI-referred visitors would be low-quality drive-bys. Adobe's measurements across trillions of US retail site visits say the opposite — and the trend flipped fast:
In March 2025, AI-sourced visitors converted 38% worse than other traffic. Twelve months later they converted 42% better — a record high (Adobe Analytics via TechCrunch, April 2026). Over the 2025 holiday season, AI-driven revenue per visit was up 254% year-over-year. And once they land, these visitors dig in:
They arrive pre-researched — the agent already did the comparison work — which is exactly why being in the agent's answer matters more than being in a list of ten links.
Why owners should care: the funnel inverts
Your funnel assumes a human: they see an ad, land on a page, get persuaded by design and copy, maybe abandon, get retargeted. Agents break almost every stage of it.
- Discovery stops being a ranked list. An agent doesn't return ten blue links for you to fight over — it returns an answer, often with two or three products. Being #4 used to mean fewer clicks; in an agent answer it means you don't exist.
- Your best salesperson can't talk to agents. Popups, urgency banners, hero videos, upsell flows — invisible to an agent parsing your product data. What persuades an agent is structured facts: price, specs, availability, shipping cost, return policy, review evidence.
- Comparison becomes relentless and literal. Agents compare every option on the criteria the shopper stated, every time. Vague shipping policies, prices hidden behind "add to cart," inconsistent spec formats — these were friction for humans; for agents they're disqualifiers.
- Loyalty gets more valuable, not less. When the agent shortlists three near-identical options, the customer who says "just order from the brand I used last time" is the one you keep without paying for. Repeat relationships and first-party data are the assets agents can't commoditize.
Why marketers should care: the channel mix is shifting under you
- AEO joins SEO. Answer Engine Optimization — being cited by AI answers, not just ranked in search — is now a real channel. It rewards different content: direct answers to real buyer questions, verifiable claims, structured data, authoritative mentions across the web. (It's why this blog exists at datajar.co/blog and answers questions in the first paragraph.)
- Paid discovery gets partially bypassed. Agents synthesize from organic, structured, and review data. The ad budget that intercepted humans mid-search doesn't intercept an agent. Watch your mix: as AI-referred traffic grows, the ROI story of every channel changes.
- Measurement needs new questions. AI-referred visitors already behave differently — they arrive pre-researched, closer to decision. You'll want to know: how much of my traffic is AI-referred? Does it convert differently? Which products do agents send people to? Your dashboards weren't built with these columns.
Six things to do now (none require betting the company)
- Make your product data machine-perfect. Complete schema.org markup, accurate titles and specs, live prices and stock in your feed. This is the single highest-leverage move — it's what agents actually read.
- State the deal-breakers in crawlable text. Shipping costs and times, return policy, warranty — on the page, in text, not buried in a PDF or an image. Agents disqualify on missing information.
- Build answer-shaped content. One page per real buyer question, direct answer in the first paragraph. This is AEO, and right now most niches have no incumbent worth displacing.
- Treat reviews as agent input. Authentic, detailed reviews are evidence agents cite when recommending. Volume matters less than substance.
- Defend the repeat relationship. Email/SMS lists, loyalty mechanics, subscriptions — the parts of your revenue no agent re-opens for bidding every time.
- Watch your data for the shift. Track AI-referred sessions, their conversion rate, their AOV, which products they land on — monthly, starting now, so you see the curve while it's still early.
That last one is where most stores will stall: these are new questions, and nobody has a dashboard for them. This is exactly the gap Datajar closes — connect your store and ask in plain English: "How do visitors from AI assistants convert compared to Google traffic?", "Which products get AI-referred traffic?" When the channel mix shifts, the stores that notice first win the transition.
The honest timeline
Agent-completed purchases are a small share of e-commerce today, and nobody serious will tell you exactly how fast that changes. But the protocols are shipped and open-sourced, every platform that matters is building on them, and the discovery layer has already moved — Adobe's numbers above are measurements, not forecasts. The move isn't to panic; it's to become the store that's easy for agents to read, verify, and recommend — before your competitors do.
Want to see how AI-referred traffic already behaves in your store? Connect it to Datajar and ask — free to start, no credit card.
Sources
- Adobe Analytics — generative-AI-powered shopping rises with traffic to US retail sites (holiday 2025: traffic +693% YoY, revenue per visit +254%, conversion of AI referrals +31%)
- TechCrunch — AI traffic to US retailers rose 393% in Q1 2026; March 2026 AI traffic converted 42% better than non-AI (vs. 38% worse in March 2025)
- OpenAI — Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol (live with US Etsy sellers; 1M+ Shopify merchants coming)
- Stripe — Instant Checkout in ChatGPT and the open Agentic Commerce Protocol, co-developed with OpenAI
- Adobe — AI-driven traffic surges across industries; retail sees biggest gains
Frequently asked questions
- What is agentic commerce?
- Agentic commerce is shopping where an AI agent — not a human clicking through pages — does part or all of the buying journey: understanding intent, researching options, comparing products, and increasingly completing the purchase. Examples include shopping answers inside ChatGPT and Perplexity, assistants like Amazon's Rufus, and checkout protocols that let agents buy directly from a merchant's catalog.
- Is agentic commerce actually happening, or is it hype?
- It's measurable now. Adobe Analytics tracked AI-sourced traffic to US retail sites growing 693% year-over-year in the 2025 holiday season and 393% in Q1 2026 — and by March 2026 that traffic converted 42% better than non-AI traffic. On the checkout side, ChatGPT users can already buy from US Etsy sellers in chat via the OpenAI–Stripe Agentic Commerce Protocol, with over a million Shopify merchants next. The share of total purchases is still small; the direction is not in question.
- How do I get my products recommended by AI agents?
- Agents choose from what they can read and verify: clean structured product data (schema.org markup, accurate feeds), clear prices, shipping and return policies stated in crawlable text, authentic reviews, and content that directly answers buyer questions. Stores that are easy for machines to parse and trust get surfaced; stores whose key information lives in images or JavaScript-only widgets get skipped.
- Does agentic commerce kill brand marketing?
- No — it changes where brand pays off. When an agent shortlists three products, a shopper who already knows and trusts your brand tells the agent to pick you; brand becomes the tiebreaker inside the agent's answer. What weakens is interruption-based discovery (paid ads shown to humans browsing), because the agent doesn't see your ads — it reads your data.