Agentic commerce
Prepare your store for AI shopping: a 90-day plan
Updated
Microsoft now tells retailers to do three things: get discovered by AI agents, get chosen by them, and understand what's working. Its playbook is built for Copilot and Bing. This guide applies the same three moves to a Shopify store whose shoppers are also on ChatGPT, Perplexity, Gemini and Claude, and lays them out over 90 days.
What is agentic commerce?
Shopping where an AI assistant or agent does part of the work a shopper used to do: finding candidate products, comparing them, checking price and stock, and increasingly completing the purchase. The shopper asks ChatGPT, Copilot, Perplexity or Gemini what to buy and acts on the answer. For a store, the question changes from where do I rank to am I the store the agent recommends, and does it have my facts right.
What does Microsoft's agentic commerce playbook recommend?
Three moves. Get discovered, with an AI-ready product feed in Microsoft Merchant Center so agents can identify and trust what you sell. Get chosen, through AI-native ads on Bing and Copilot, Copilot Checkout so a shopper can buy inside the conversation, and a Brand Agent on your site. Understand what's working, with AI Visibility insights in Microsoft Clarity. It comes with a 90-day activation roadmap that spends the first half on feed and checkout foundations and the second on content and measurement.
Does the playbook cover ChatGPT, Perplexity and Gemini?
No. It is built around Microsoft's surfaces: Copilot, Bing, Merchant Center and Clarity. The three moves apply everywhere, but the tools it names only measure and act on Microsoft's share of AI shopping. A store that wants to be chosen on ChatGPT, Perplexity, Gemini and Claude as well needs to make its product facts readable to all of their crawlers and measure its visibility on each engine directly.
The three moves, for a store on every engine
Microsoft's wording is in the first column. The second is what the same move means when the shopper is asking ChatGPT or Perplexity instead of Copilot. The third is where PositionBird fits, and where it does not.
| Move | What Microsoft says | What it means on Shopify | What PositionBird measures |
|---|---|---|---|
| Get discovered | An AI-ready product feed in Microsoft Merchant Center, so agents can identify and trust what you sell. | Every engine needs the same thing the feed gives Copilot: product facts an agent can read and trust. On Shopify that is complete titles, specs, price, availability and policies in the page text and in Product structured data, with AI crawlers allowed in robots.txt. | Site Health checks structured data, FAQ content and crawler access per product page. The scan shows whether each engine can find you at all. |
| Get chosen | AI-native ads across Bing and Copilot, Copilot Checkout so a shopper can buy without leaving the conversation, and a Brand Agent on your site. | Outside Microsoft's ad surfaces, being chosen is organic: the engine names you in its answer, at a rank, with the right price and stock. Shopify handles the checkout surfaces; your job is to be the store the answer recommends. | ShopBird asks each engine real buying questions and records whether you are recommended, at what rank, which competitor wins instead, and whether the quoted price and stock are right. |
| Understand what's working | AI Visibility insights in Microsoft Clarity: which AI platforms cite your brand, competitor comparisons, and how AI-referred shoppers behave after the click. | Clarity sees Copilot and Bing well and the rest as referrers. Shopify's AI sales dashboard (added at Summer Editions, June 2026) now reports orders from ChatGPT, Gemini, Copilot and Shop, but not engine-by-engine visits or Perplexity and Claude, so the rest still has to be counted by referrer domain by hand. | The citation ledger records which sources each engine trusted and whether you were among them. RevenueBird credits orders to the AI engine that started the visit, first touch, and reports it weekly. |
The 90-day plan
Same shape as Microsoft's roadmap: foundations first, content and measurement second. Each phase produces something you can show someone, and the first and last phases are the same scan run twice.
- 01
Take a dated baseline before you change anythingDays 1 to 7
Write 10 to 25 buyer-intent prompts shoppers would ask about your category and list 3 to 5 competitors. Run them through the engines and record, per prompt and engine, whether you were named, who was named instead, and which sources were cited. A free scan gives you this in about a minute. Without a baseline, nothing you do in the next 83 days can be shown to have worked.
- 02
Make the catalog readable to agentsDays 8 to 30
This is Microsoft's feed-readiness step, applied to your own pages. Confirm robots.txt allows OAI-SearchBot, PerplexityBot, Claude-SearchBot and Bingbot. Put price, availability, materials, sizing and shipping policy in page text, not only in images or tabs that need a click. Add Product and Offer structured data on every product page and FAQ structured data where you answer real questions. Fix the pages your scan flagged first.
- 03
Check the facts AI quotes about youDays 31 to 45
An agent that finds you but quotes the wrong price, says you are out of stock, or invents a return policy loses the sale as surely as one that never names you. Pick the ten facts that matter most to a buyer and verify what each engine says about them. Correct the source pages the engines cited, since that is where the wrong fact came from.
- 04
Earn the citations you are missingDays 46 to 75
Microsoft's second half is content. In the baseline you saw which sources each engine trusted for your prompts: review profiles, category roundups, forum threads, marketplaces. Work the list. Claim or complete the review profile, pitch the roundup that names your competitor and not you, answer the forum question. Off-site sources carry more weight than your own blog.
- 05
Count the visits and the ordersDays 46 to 75, in parallel
Set up a saved report in Shopify for AI referrer domains and, if you advertise with Microsoft, Clarity's AI Visibility report. Credit the AI visit that came first, not the last click. Keep AI sessions and AI orders as separate lines, because a small session count with a high order rate is the normal pattern for this channel.
- 06
Re-scan, compare, and set the cadenceDays 76 to 90
Run the same prompts on the same engines and put the two scans side by side: prompts gained, prompts lost, competitors overtaken, facts corrected. Then pick the scan schedule you can act on, weekly for most stores, and turn on alerts for a lost mention or a competitor taking a prompt you held. The plan ends; the loop does not.
I am on Shopify. What is my equivalent of an AI-ready product feed?
Your product pages and their structured data. Agents that are not reading a Merchant Center feed read the page: title, description, price, availability, variants, shipping and returns, plus the Product and Offer schema behind it. Shopify generates llms.txt for every store, but there is no evidence engines read it, so the page itself has to carry the facts. Keep them in text an agent can parse, not in images or collapsed tabs.
Do I need agentic checkout to benefit?
Not to start. Agentic checkout, where the purchase completes inside the assistant, only matters once an agent recommends you, and today most AI-referred orders still arrive as a visit to your store. Microsoft's data says those visitors convert 42% better than non-AI traffic, so being recommended with the right facts is the step that pays first. Checkout surfaces come from your platform and the engines; being chosen is yours to earn.
How do I measure AI-driven traffic and visibility?
Two halves. Traffic you can see in your own analytics by filtering referrer domains like chatgpt.com, perplexity.ai and copilot.microsoft.com, and in Clarity's AI Visibility report if you use Microsoft Advertising. Visibility you cannot see from your store at all, because the answers that named a competitor sent you no visit. That half has to be collected from the engines on a schedule, which is what PositionBird does, and a free scan gives you a dated starting point today.
What the data says
Five things from primary sources, each dated. The first four are Microsoft's own; the last is Shopify's.
- 3 moves, 90 days is the shape of Microsoft Advertising's playbook, Prepare for the Age of AI Shopping: get discovered with an AI-ready product feed in Microsoft Merchant Center, get chosen through AI-native ads, Copilot Checkout and a Brand Agent, and understand what's working with AI Visibility insights in Microsoft Clarity, with a 90-day activation roadmap. Microsoft Advertising, Prepare for the Age of AI Shopping, 2026-10-02
- 72% of U.S. consumers expect AI-assisted shopping experiences from retailers within the next year, in Microsoft Advertising's own consumer survey of 1,028 people run in August and September 2025. Microsoft Advertising, Rebuilding the Web FY26 consumer research, cited in the playbook, 2026-10-02
- +42% better conversion for AI-referred visitors than non-AI traffic, which the playbook attributes to Adobe Analytics' Q1 2026 AI traffic report, with the line that AI-referred visitors arrive ready to buy. Adobe Analytics, as cited by Microsoft Advertising, 2026-10-02
- 4 foundations in Microsoft's holiday follow-up: feed readiness so AI can discover more of your catalog, measurement readiness to connect interactions to outcomes, campaign readiness, and AI readiness to help AI find and understand your brand. It tells merchants to set up Clarity to baseline AI-referred visitors against other sources. Microsoft Advertising blog, Get your brand AI-ready for the holidays, 2026-10-02
- 3x year-over-year growth in AI-driven traffic and orders across Shopify merchants in the second quarter of 2026, with AI-referred sessions landing directly on a product page 2.5 times as often as traditional search. Shopify Q2 2026 earnings call, 2026-08-05
PositionBird covers the discovery and measurement halves of this plan across every engine, not one. The scan asks ChatGPT, Perplexity, Gemini and Claude your shoppers' real buying questions through their official APIs and records whether you are named, who is named instead, which sources were trusted, and whether the price and stock quoted match your store. Site Health checks the product-page facts agents read. The RevenueBird pixel credits orders to the AI engine that started the visit. What it does not do is sell you ads or run your checkout; those stay with your platform and the engines. For the weekly routine after day 90, see AI search monitoring for ecommerce, and for counting the visits in Shopify's own reports, see how to see ChatGPT traffic in Shopify.
Day 1 of the plan is a free scan
No account. Whether AI agents recommend your store today, which competitor they name instead, and what to fix first.