Shopify has opened a new sales surface. With Agentic Storefronts, merchants can manage how their products appear across ChatGPT, Microsoft Copilot, Google AI Mode, Gemini, and other AI shopping experiences from Shopify Admin. Your products can now show up when someone asks an AI what to buy.
That sounds like free distribution, and in a sense it is. But being in the catalog and being the answer are very different things. When a shopper asks for “a linen duvet cover that survives a dog and a toddler,” ChatGPT recommends two or three products. Not two or three hundred. The question that matters is how it picks, and whether it picks you.
The numbers say this is worth your attention now, not next year. According to Shopify’s Q1 2026 commerce data, referral sessions from AI chatbots grew more than 8 times year over year and AI-attributed orders grew nearly 13 times. More than half of those sessions begin on product pages; among product-page landing sessions, AI referrals convert nearly 50% better than organic search and carry a 14% higher average order value. This channel has moved from curiosity to measurable revenue.
We’ve written before about how answer engines choose sources and what llms.txt does and doesn’t do. This post is the practical layer on top: what to actually change on your store, in what order. Think of it as an audit you can run this month.
Step one: find out where you stand
Before you optimize anything, get a baseline. It takes an evening.
Open ChatGPT and Perplexity and ask them the five questions a real customer would ask before buying from you. Not “tell me about [your brand].” The real ones. “Best cold brew maker under $50.” “Gentle face wash for rosacea that ships to Chennai.” Write down who gets recommended, who gets cited, and where you show up, if at all.
Then open Shopify Analytics and filter your traffic by referrer. Shopify now breaks out AI sources like ChatGPT, so you can see how many sessions and orders already come from assistants. Most founders who do this are surprised in one direction or the other. Either the channel is already producing quiet revenue, or competitors own every answer. Both are useful to know.
Keep the list of prompts. You’ll re-run them monthly, and that’s your scoreboard. There’s no rank tracker for this yet that we’d trust more than your own eyes.
Fix the product pages before anything clever
Here’s the unglamorous truth about AI recommendations: assistants recommend products they can understand and defend. If your product page can’t answer a customer’s question, it can’t answer the AI’s either.
Shopify’s own guidance on this is blunt. Make pages fact-rich, in simple declarative language. Dimensions, materials, certifications, care instructions, compatibility. Their example is Brooklinen, which lists exact dimensions for every size and detailed material specs on each product. An assistant can lift “480 thread count, long-staple cotton, fits mattresses up to 15 inches deep” and build a recommendation on it. It can’t do anything with “luxuriously soft bedding you’ll love.”
Two specific edits pay off fastest:
Answer the buying question in the first few sentences. Whatever a customer most needs to know before purchasing, say it plainly at the top of the description, not in paragraph four. Assistants quote what’s clear and easy to extract. So do skimming humans, which is why this edit tends to lift conversion on its own.
Add a real FAQ to key pages. Real meaning the questions your support inbox actually gets, answered honestly. “Does it work with induction stoves?” “How long does shipping to Bangalore take?” Skip the padded corporate version. One honest paragraph per question beats five vague ones, and FAQ content is exactly the shape assistants like to cite.
Get your structured data straight
Structured data is JSON-LD markup that hands machines your price, availability, ratings, and specs in a format they don’t have to guess at. It’s been table stakes for Google rich results for years. Now it feeds the AI layer too.
On Shopify the news is mostly good. Decent themes ship with Product schema out of the box. But “mostly” is doing some work in that sentence. Run a few of your product pages through Google’s Rich Results Test and look for what’s missing. The common gaps we find on audits: review schema absent even though the store has hundreds of reviews sitting in an app, FAQ schema missing from pages that have FAQ content, and price or availability fields silently broken by an old theme customization.
If your reviews live in an app, check whether that app injects review schema. Some do it well, some don’t do it at all. This is one of the few places where a store can get meaningfully more quotable with a day of technical work and zero new content.
Be known somewhere other than your own site
This is the part most stores skip, and it’s probably the most decisive.
Assistants don’t take your word for it. When one recommends a product, it leans heavily on what the wider web says: reviews on sites it trusts, mentions in publications, comparisons on forums. Your own product page supplies the specs. Third parties supply the credibility.
Shopify’s guide suggests a few concrete moves here, and they match what we see working. Find out which publications and sites keep appearing as sources when AI answers questions in your niche. Ask the assistants your customer prompts and note the citations; that’s your target list. Then earn your way onto it, through PR, affiliate relationships, or simply having a product good enough to review.
If you have data nobody else has, publish it. Original research makes you a primary source, the thing every assistant wants to cite. A pet food store that surveys 500 dog owners about feeding habits owns a citation magnet no competitor can copy-paste.
None of this is fast. That’s exactly why it works. The stores getting recommended today started earning mentions before it was obvious they should.
Check the plumbing
A few mechanical things round out the audit.
Since Agentic Storefronts surface your catalog directly inside ChatGPT, Google’s AI Mode, Copilot, and Gemini, your catalog data itself is now customer-facing in a new way. Clean titles, accurate variants, current availability, filled-in shipping and returns policies. An agent reading your store makes decisions on exactly these fields, and a stale “out of stock” or an empty returns policy costs you recommendations you’ll never know about.
Then there’s llms.txt, the small markdown file that hands AI a curated map of your store. We covered it in depth here, including the honest caveat that major crawlers haven’t committed to it. Short version: cheap bet, real option value, don’t mistake it for the whole job.
And keep the store fast. Assistants retrieve and read your pages live, while the shoppers they refer often arrive with high intent. A page that makes crawlers struggle—or makes a ready-to-buy customer wait—quietly loses the opportunity. If your speed situation is murky, start here. If you want the product-page implications of Shopify’s latest data, read our analysis of AI-referred shopping behavior.
What this looks like as a plan
Run the baseline this week: five customer prompts in two assistants, plus the AI referrer report. Spend the next two weeks on your ten best-selling product pages, making each one fact-rich with a real FAQ. Validate structured data and fix the gaps, one focused day of work. Then start the slow game of earning third-party mentions, one outreach or one piece of original research a month.
Re-run your prompts every month and watch the referrer report. When you start showing up in answers, you’ll see it in both places.
The founders who win this channel won’t be the ones who found a trick. There isn’t one. They’ll be the ones whose stores were genuinely the easiest to understand, verify, and recommend, back when most competitors were still treating AI shopping as someone else’s problem.
PixelSplash designs, builds, and scales Shopify storefronts for brands that have outgrown their theme. Want an AEO audit of your store, from schema to product copy? Book a free consult. 30 minutes, zero pitch.