AI shopping agents: a retail toolkit for business students
AI traffic to US retail sites is forecast to jump 130%

AI traffic to US retail sites is forecast to jump 130% this holiday season.
that's Adobe's number, built on more than 1 trillion retail site visits (so not a vibe).
AI shopping agents are becoming a real sales channel, and most business students still treat them as a chatbot feature.
here's the toolkit we'd actually learn, and how to use it without getting fooled.
What's actually happening with AI shopping agents
the last few weeks in agentic commerce:
→ Adobe forecasts $275.1 billion in US online holiday sales (up 6.7%), with AI traffic to retail sites up 130% and the biggest jump on Thanksgiving (159%) → Forter launched an Agentic Commerce Engine on 13 October, saying agentic orders across its network more than quadrupled from August to October... and automated fraud doubled in the same window → Google's Universal Commerce Protocol, co-developed with Shopify, Etsy, Wayfair, Target and Walmart, powers checkout inside AI Mode and Gemini, with the retailer staying seller of record
and two things that didn't go to plan:
- OpenAI scaled back Instant Checkout in March. Shopify's "agentic storefronts" now send most buyers back to the merchant's own checkout.
- Kroger's AI shopping assistant launched with ads already switched on, labelled as sponsored. Walmart announced ad features for its Sparky assistant in January.
so the agent that "helps you shop" can also be a billboard. worth knowing before you build a career on it.
What uni won't teach you about agentic commerce
most marketing and retail modules still assume a human walks the funnel: sees an ad, lands on a page, adds to basket.
with AI shopping agents:
→ the "visitor" might be software reading your product data, not your hero banner → the shelf is an answer, not a search results page with ten slots → the customer relationship can sit with Google or OpenAI, not with you
case studies in class are usually two years old. this channel didn't exist two years ago.
The toolkit: 4 tools worth your time
- three assistants, one basket. use ChatGPT, Gemini and one other assistant as test shoppers. ask the exact same question ("best running shoes under £100 for flat feet") and screenshot the answers. which brands show up, which retailers get the click, and does anything look sponsored?
- product feeds. open a free Google Merchant Center account (or read its docs) and look at what a product listing actually needs: titles, attributes, stock, price. agents read structured data, so this is the new shop window.
- referral tracking. in Google Analytics, check the referral report for traffic from AI assistants on any site you can access (a society website, a friend's shop). no access? read Adobe's methodology instead and learn how "AI traffic" gets measured.
- the protocol announcements. read Google's UCP post and the Modern Retail piece on OpenAI's pullback, back to back. one is the pitch, the other is what happened when merchants actually tried it.
The critical-thinking move: 5 questions for any AI shopping agent
run any agentic commerce tool or headline through these:
- who does the agent work for? the shopper, the retailer, or whoever paid for the sponsored slot? Kroger launched with ads on day one, so this isn't hypothetical.
- where does checkout happen, and who owns the customer after? the seller of record, the returns, the email list. Google keeps the retailer as seller of record, but the checkout happens inside Google.
- forecast, measurement or vendor stat? Adobe's 130% is a forecast. Forter's "quadrupled" comes from its own network, and it sells the fix. none of these are wrong, they're just different strengths of evidence.
- what breaks when the customer is a bot? fraud, returns abuse, price scraping. if agentic orders grow, the attack surface grows with them.
- would a real shopper trust this answer? the OpenAI pullback is a reminder that people (and merchants) don't always follow the demo.
What to do this month
- run the basket test on three products in a category you care about. put it in a one-page doc with screenshots. honestly, that alone beats most coursework on AI in retail.
- pick one retailer and check whether its products appear in AI Mode or ChatGPT answers. write down what's missing from its listings.
- save Adobe's forecast now and compare it with Adobe's actual holiday numbers in January. forecast vs reality is the best free lesson in reading industry stats.
- write a 300-word memo answering question 2 above for one brand. that's the kind of thinking an ecommerce team actually needs.
so TLDR: AI shopping agents are real traffic now, but the agent isn't always on the shopper's side. learn the tools, then ask who's paying.
if your favourite brand vanished from every AI shopping answer tomorrow, how long before you'd even notice?
Sources
- Adobe projection: 2026 online holiday sales to reach $275.1 billionDigital Commerce 360 · 29 SEP 2026
- Forter just shipped the most complete merchant-side stack for AI agent traffic. Holiday peak is five weeks awayForkast · 13 OCT 2026
- New tech and tools for retailers to succeed in an agentic shopping eraGoogle · 11 JAN 2026
- Shopify says purchases are coming inside ChatGPT through agentic storefronts as OpenAI retreats on Instant CheckoutModern Retail · 12 MAR 2026
- Kroger's AI shopping assistant has ads at launchModern Retail · 04 AUG 2026
Get the next issue by email
theVANTAGEd is written by students to bring critical thinking to AI across real industries. One issue every two weeks.