How AI is changing the way people shop, and what your Shopify store should do about it

For twenty years the path to a purchase started with a search box and ten blue links. That path is splitting. A growing share of shoppers now open ChatGPT, Gemini or Perplexity, describe what they want in a sentence, such as "a breathable running top for hot weather under $60 that ships to Dubai", and get back a short list of specific products, with reasons. Google has moved the same way with AI Overviews and AI Mode. The assistant does the browsing; the shopper does the choosing.

That changes what a Shopify store has to be good at. You are no longer only competing for a click on a results page. You are competing to be understood and recommended by a machine that has read your pages, your reviews and everyone else's. Here is what that means in practice, and the order we work through it on our clients' stores.

How the assistants actually choose products

Nobody outside the labs knows the exact ranking rules, and they change often. But the pattern is consistent across the tools we test with: an assistant will only recommend a product it can identify clearly, describe confidently and justify. It needs to know what the thing is, who it's for, what it costs, whether it's in stock, and why other people liked it. If any of that is missing or contradictory on your page, the assistant quietly picks the competitor whose page answered the question.

So the work is less about tricks and more about removing every reason a careful reader would hesitate. The good news: the same fixes that help an AI assistant also help Google, and they help a human on a phone at 11pm.

1. Give the machines structured product data

Shopify themes ship with basic Product schema, but "basic" is the problem. Check a product page with Google's Rich Results Test and look for what is missing. On most stores we audit it's some mix of:

• Offer details: price, currency, availability, and a real priceValidUntil if you run sales.
• Identifiers: brand, SKU, and GTIN or MPN where you have one. Identifiers let an assistant match your listing to the product it already knows about.
• Aggregate rating and review markup, pulled from your review app, not typed by hand.
• Shipping and returns: Google's shippingDetails and hasMerchantReturnPolicy properties. Assistants increasingly quote delivery time and return windows in their answers.

If your theme can't output these, a small theme customisation or a well-chosen SEO app will. Then keep your Google Merchant Center feed clean and complete; several assistants lean on merchant feeds as a trusted source of price and stock.

2. Write product pages that answer the question

Read your best-selling product page and ask: if a stranger asked "is this right for me?", does the page answer in the first screen? Assistants extract facts from plain prose, so make the facts plain:

• Say who it's for and what it's for in the first two sentences. "Sticky leggings for pole and aerial training, with grip fabric on the inner leg and no seam at the knee."
• Put materials, sizes, dimensions, care and compatibility in a scannable spec list, not buried in a paragraph.
• Add a short FAQ with the questions your customer service actually gets (sizing, shipping to specific countries, how it compares to your other model) and mark it up with FAQPage schema.
• Use the words customers use, not only your brand's names for things. Assistants match intent to language.

3. Make your reviews readable by people and machines

Reviews are the evidence an assistant cites when it says "customers mention the fit runs small." Stores with no reviews, or reviews locked inside an image carousel, give it nothing to cite. Collect reviews after delivery (email and WhatsApp both work well), show them as real text on the product page, and make sure the review app exposes them in schema. A handful of honest, specific reviews beats a hundred five-star "great!"s.

4. Decide who is allowed to read your store

Every AI assistant crawls the web with its own bot: GPTBot, PerplexityBot, Google-Extended and others. Some Shopify stores block them in robots.txt, sometimes on purpose and often by accident after installing a "block bots" app. Blocking is a legitimate choice for content businesses; for a store that wants to be recommended, it's like taking your products off the shelf. Open your robots.txt.liquid, read what is blocked, and make it a decision rather than a default.

5. Speed, consistency and the boring foundations

Assistants fetch pages on a budget. Slow, script-heavy product pages sometimes never get fully read. Keep image weight down, remove apps you no longer use, and make sure the same product has the same name, price and description everywhere it appears: your store, your feed, your marketplace listings, your social shop. Inconsistency reads as untrustworthy, to a model and to a person.

What not to do

Don't stuff pages with "best", "top rated" and keyword lists. Assistants are good at spotting it, and it damages the human read. Don't invent reviews. Don't create fifty thin pages for every phrase variation. And don't rebuild your whole store for a channel that is still settling; make the pages you already have unambiguous first.

Where to start this week

1. Run three product pages through the Rich Results Test and list what's missing.
2. Rewrite the first two sentences of your top five products so a stranger knows who they're for.
3. Check robots.txt for blocked AI crawlers and decide on purpose.
4. Set up a post-delivery review request if you don't have one.
5. Do the "quick test" above and write down the gap.

None of this is glamorous. All of it compounds. The stores that get recommended a year from now will be the ones whose pages were simply the clearest answer to the question.