Use AI for three jobs. Let AI tools find your products, let an agent on your own site help people buy, and let AI handle the copy, support and reports behind the counter. All three run on the same thing, your product data. Fix that first. In September 2026, shoppers sent by AI converted 90% better than other visitors.
I've spent fifteen years designing and marketing online products. The web used to have one reader, a person with a browser. Now it has two readers, people and AI. A store has two shoppers in the same way. One is a person. The other is an AI that reads your catalog for that person, compares you with everyone else and sends over a shortlist.
This guide covers what AI does for a store, what the data says, how to get ready, and which tools to use. Every number has a source at the end.
What does AI actually do for an online store?
It does three jobs, and each one is a different project.
Three jobs for AI in a store, all fed by one catalog
AI finds you. ChatGPT, Gemini, Copilot and the newer personal shopping agents read product feeds and pages, then recommend products in a conversation. This is agentic commerce. The shopper asks, the AI shortlists, and the click lands on a product page.
AI sells for you. A shopping agent on your own site answers questions, builds a cart and hands it to your checkout. Anthropic says retailers running shopping agents on Claude see carts up to 35% larger and shoppers 60% more likely to complete a purchase.7 That's the vendor's own figure, with no published method, so treat it as a ceiling, not a forecast.
AI runs the store. Product copy, image alt text, support replies, weekly reports, pricing checks and automations. Here the evidence is independent. In an NBER study of 5,179 support agents, an AI assistant raised issues solved per hour by 14%, and by 34% for the least experienced staff.9
The order of those jobs matters less than the thing they share. All three read the same titles, attributes, prices, stock and policies. Weak product data means AI can't recommend you, your agent gives wrong answers, and your automations copy the mistakes.
Do shoppers really buy through AI?
Yes, and they buy better than average. The volume is still small.
AI-referred shoppers vs everyone else, US retail, September 2026
Adobe tracks more than a trillion visits to US retail sites. In September 2026, visits from AI tools converted 90% better than other visits and brought 83% more revenue per visit. A year earlier, in August 2025, AI visits converted 9% worse.1 AI traffic to US retail sites grew 71% in the same twelve months.1
Shopify sees the same pattern on its own merchants. On its Q3 2026 call, it said AI traffic to stores was up 2.6 times and orders from AI were up 3.3 times on a year earlier. AI channels sent shoppers straight to product pages at three times the rate of organic search, and those shoppers converted 64% more often.4
Shopify was careful to add that the volume is still small.4 So keep two facts in your head. AI is a minority of your traffic today. It's also the best traffic you get, and it's growing faster than any other channel.
One more number from the call matters for small stores. 74% of AI-attributed purchases came from outside Shopify's top 100 product categories.4 AI rewards the long tail. A niche product with clear data can win a recommendation against a big brand with vague data.
Did AI checkout replace your store?
It tried, then came back. This is the part most guides get wrong.
Checkout moved into chat, then back to your store
In September 2025, OpenAI launched Instant Checkout, so people could buy inside ChatGPT.14 In January 2026, Shopify and Google launched the Universal Commerce Protocol, an open standard for agents to shop with any merchant, backed by more than 20 retailers and platforms.6 Then, on 24 March 2026, OpenAI said the first version of Instant Checkout wasn't flexible enough. It now lets merchants use their own checkout, and ChatGPT focuses on product discovery.5
Anthropic's commerce blueprint, released in September 2026, makes the same bet from the other side. Its shopping agent lives inside the merchant's own site and leaves payment to the merchant.7 I explain the two standards behind all this in agentic commerce protocols.
So here's where things stand in late 2026. AI does the finding and the comparing. Your store still does the selling, the payment and the customer relationship. That's good news if you own your site and your data, and bad news if your product pages can't be read.
What's my take on AI in ecommerce?
The catalog is the storefront now.
For twenty years, ecommerce design meant the page a person sees. The hero image, the layout, the buttons. That page still matters, because the click from AI lands on it. Before the click, though, the AI only sees your data. It reads the product title, the attributes, the price, the stock, the shipping rules and the returns policy. If those are missing, locked in images or only drawn by JavaScript, the AI skips you.
How much of a product page AI can read, July 2026
Most stores aren't ready. Adobe's first benchmark, in April 2026, found that AI could read only 66% of an average US retail product page.3 By July, the sector spread was wide. Sporting goods product pages scored 70%, and general merchandise only 48%.2
Store owners are noticing that the ground has moved. I pulled the US search data for this guide from Google Keyword Planner. Searches for Claude with Shopify and Claude with ecommerce went from almost nothing to the same band as ChatGPT with Shopify. Meanwhile, the ChatGPT and OpenAI searches fell back after their 2025 peak.
Store owners now search for Claude, and less for ChatGPT
My read is simple. People have stopped asking which chatbot will sell for them. They're asking which AI will help them build and run the store. That's a better question, and it's the one this guide answers.
There's a second reason I like where this is heading. It puts people and AI on the same side. The AI does the reading and the sorting. The person makes the choice. A good store makes both jobs easy.
How do you get your store ready for AI shoppers?
Work through six steps, in this order. Steps 1 to 3 feed every AI channel at once.
Six steps to a store AI can sell
- Check what AI can read. Paste a product page into the AI shopping readiness checker. It reads the raw page the way a crawler does and scores the product data, schema, policies and crawler access.
- Fix the product data. Write titles the way people describe the product: brand, product type, key attribute. Fill every attribute your platform offers, such as size, material, compatibility and GTIN. Put price, stock and delivery times in text, never only in images.
- Add schema and a feed. Mark up each product page with product data for AI shopping, including the offer, shipping and returns. The product schema generator writes the code for you. Keep your Google Merchant Center feed clean, because Google's agents read it.
- Switch on the AI channels. On Shopify, Agentic Storefronts let you set up your data once and show it across the AI channels Shopify connects to.6 On other platforms, submit a product feed to each channel you want, and let the AI crawlers in through your robots.txt file.
- Keep checkout and policies clear. AI now hands shoppers to your own checkout.5 Returns, shipping costs and delivery times must be easy to find in text, because AI quotes them when it compares you with a rival.
- Put AI behind the counter, then measure. Add a shopping agent or AI support only once the data is right. Tag AI referrals in your analytics so you can compare their conversion with search and email.
The steps above are about what AI reads. For what people see on the page, run the product page audit and read ecommerce website design that sells.
If you sell services as well as products, the same rules apply to your service pages. I cover them in how to get cited by ChatGPT.
Should you use ChatGPT, Claude or Shopify's own AI?
Use more than one, for different jobs. Here's how the main options compare in November 2026.
| What it's for | Where it runs | Best for | |
|---|---|---|---|
| ChatGPT shopping | Being found and compared | ChatGPT, via your product feed | Discovery, from any platform |
| Google AI Mode and Gemini | Being found, and checkout through UCP | Google, via Merchant Center | Stores already in Google Shopping |
| Shopify Sidekick and Canvas | Running and building a Shopify store | Shopify admin | Shopify merchants who want it built in |
| Claude with Shopify's AI Toolkit | Building, editing and analysing the store | Claude Code, connected to Shopify | Custom work without a developer on call |
| Claude commerce agents | A shopping agent on your own site | Your site, on Claude | Brands that want their own agent |
| Medusa with Bloom | Building a custom open-source store | Your own code and hosting | Teams that need full control |
ChatGPT is now mainly a discovery channel. OpenAI expanded its Agentic Commerce Protocol so merchants can share product feeds and promotions, and ChatGPT shows richer product comparisons.5 You don't build anything in ChatGPT. You make sure your feed and pages are good enough to be picked.
Shopify's own AI is the easiest start if you're on Shopify. That's what most people mean when they search for Shopify AI. Sidekick created around two-thirds of all Shopify Flow automations in Q3 2026. Canvas, launched the same quarter, builds a custom storefront from a description.4
Claude is where store owners go for work Sidekick can't do. Shopify's Dev MCP server, shipped in its Winter '26 Edition, connects AI coding tools like Claude Code to Shopify's documentation and APIs.13 In practice, you describe a change in plain words, Claude writes and checks the code against Shopify's schema, and you approve it. For a shopping agent on your own site, Anthropic's blueprint gives developers working code to start from.7
Medusa is the open-source route. It's an open-source commerce engine you host yourself or on Medusa Cloud. Hosting it yourself means no platform fee on sales. Medusa ships an MCP server and agent skills for AI coding tools, and Bloom, which builds a store from a prompt.1112 It suits teams with unusual pricing, B2B rules or marketplaces that would fight Shopify's limits. If you're still choosing a platform, start with my platform comparison.
Can you build the store itself with AI?
Yes. The hard part is knowing what to ask for and checking what comes back.
Companies are already doing it. In McKinsey's 2026 survey of 1,719 organisations, 32% said they had decided against buying at least one software product or feature, because they could build it in-house with agentic coding tools.8
The tools are ready for small teams too. Shopify Canvas builds a storefront from a description.4 Medusa's Bloom builds a storefront, admin and backend from a single prompt.11 Claude Code with Shopify's Dev MCP edits a real store's code safely.13 I wrote about the risks of this way of building in vibe coding a website. The short version is that AI writes fast and checks badly. Someone on your side has to know what good looks like.
That's the work I do. I build the store with your team, using the same AI tools you'll use afterwards. Your team and I set up the product data, the schema, the design system and the checks together. Then I train your team to run and improve the store with AI on their own. There's a package for small businesses and one for enterprise teams. If you want help with the data side alone, that's my context systems service. If you want the full build, it's website design.
What does AI in ecommerce look like for an enterprise?
Bigger companies are further along, and their problem is control, not access.
In McKinsey's 2026 survey, 40% of organisations with more than $1 billion in revenue said they were scaling AI agents, up from 27% a year earlier. Smaller organisations stayed flat at 22%.8 Yet only 37% of all respondents said AI had any effect on their profits, about the same as the year before.8
So the gap in enterprise AI is rarely the model. It's the context the model gets, and the rules around what it may do. An enterprise store agent needs three things a small one can do without.
- One source of truth. Product, price and stock data from the ERP or PIM, never copied by hand into prompts. That's a context system.
- Limits on authority. Caps on discounts, refunds and cart changes, with a person approving anything above them. Anthropic's blueprint builds this in, and its merchant agent waits for a person to approve changes.7
- Measurement by channel. AI referrals, agent conversations and assisted orders tracked apart from search and email, so the board sees real numbers.
What does AI save a store each month?
It saves hours, mostly in writing and support. The evidence is solid for both.
In a 2023 study in Science, professionals using ChatGPT finished writing tasks 40% faster, and their work was rated 18% better.10 The NBER support study found a 14% gain in issues solved per hour.9 Both are averages. Your gain depends on how much of the work is routine.
The AI store time saver applies those two figures to your own numbers. Put in your new products a month, your support tickets and your reporting time. It shows the hours you'd get back and where to start.
Questions people ask
What is the best AI for ecommerce?
There isn't one. Use ChatGPT and Google for being found, your platform's AI for daily tasks, and Claude or another coding agent for custom work. Good product data matters more than the choice of model.
Can ChatGPT run my Shopify store?
No. ChatGPT can find and recommend your products, and Shopify can show your catalog there. Running the store means Sidekick inside Shopify, or a coding agent like Claude connected through Shopify's developer tools.
Do I need to change my checkout for AI shoppers?
Usually not. Since March 2026, ChatGPT sends shoppers to merchants' own checkouts. On Shopify, agents can pay through Shop Pay. Make shipping costs, delivery times and returns easy to read in text.
Is Medusa a good Shopify alternative?
For teams with developers, yes. Medusa is open source, so hosting it yourself carries no platform fee on sales. Hosting, security and upgrades are your job, though. Stores with standard needs usually do better on Shopify.
How do I know if AI can read my product pages?
Paste a product page into the free AI shopping readiness checker. It reads the raw HTML the way a crawler does. It checks the product schema, price, stock, shipping, returns and which AI crawlers your robots.txt file allows.