ChatGPT is becoming a sales channel

AI Build Your Business
25

The Intent → AI → Decision → Transaction framing is right, but there’s a step most miss between “AI” and “Decision.”

The AI has to trust you enough to recommend you in the first place.

That’s the bit that isn’t being talked about enough. Shopify giving you access to Agentic Storefronts doesn’t mean ChatGPT or Perplexity will actually surface your products when someone asks. Access and visibility are two completely different things. Right now most merchants assume being on Shopify means they’re in the game. They’re not, they’re just eligible.

What actually drives the recommendation is whether the model has enough coherent, trustworthy signal about your brand to confidently put your name in front of a buyer. That comes from third-party sources the model already trusts: Reddit threads, roundups, editorial mentions, review platforms: describing you consistently. Structured product pages. pages like llms.txt agents.md, robots.txt. Individually however, they’re not what creates the recommendation.

The stores winning in AI right now mostly aren’t doing anything sophisticated. They just happen to be mentioned in the right places, described consistently, and structured clearly enough that the model can form a confident answer.

The question merchants should actually be asking isn’t “how do I optimise for agentic commerce” but rather “does AI know who I am and what I sell, and if I asked it right now, would the answer be right?”

Most would be surprised by what comes back.

29

This topic was really interesting to me, so it pushed me to dig deeper and read several sources, especially the official Shopify and OpenAI documentation.

My main takeaway is that this is a real change in product discovery, but it probably won’t replace traditional SEO. Instead, merchants will need to combine SEO with better product data.

ChatGPT’s general web search and its product search work similarly at a high level: both try to understand the user’s intent and find relevant information. However, product search relies much more on structured and up-to-date data from sources such as Shopify Catalog—things like titles, descriptions, variants, price, availability, shipping, policies and reviews.

I also learned that there are two separate challenges:

  1. Getting your product selected as one of the recommendations.

  2. Getting your store ranked above other merchants selling that product.

Simply being included in Shopify Catalog makes a product eligible, but doesn’t guarantee that ChatGPT will recommend it.

Another interesting point is that files like llms.txt aren’t a magic ranking trick. Shopify now creates AI discovery files automatically, while Shopify Catalog remains the main source of product information.

So, for me, the practical “AI SEO” strategy is not about finding a hack. It is about keeping product information complete and accurate, clearly explaining who each product is for and which problems it solves, maintaining competitive pricing and availability, collecting genuine reviews, and continuing normal SEO.

At the moment, ChatGPT is mainly a discovery and referral channel for Shopify merchants, with customers completing checkout through the merchant’s store. But even at this stage, it could become an important source of high-intent customers.

I’d be interested to hear from merchants who are already seeing ChatGPT traffic or sales. What kinds of searches are bringing customers to your products?

30

One thing I’d add from working on buyer visibility: AI shopping is not just one problem.

There are at least three separate layers merchants need to think about:

  1. Discovery: can the AI understand what your store sells?
    This is where clean product data, titles, descriptions, availability, pricing, policies, reviews, and Shopify Catalog matter most.

  2. Decisioning: when several stores sell similar products, why should the AI recommend yours?
    This is where the “extra context” starts to matter: shipping, trust signals, return policy, bundles, loyalty/rewards, and whether the AI can explain the customer’s net value clearly.

  3. Transaction / attribution: if the customer comes from ChatGPT, Claude, Copilot, Gemini, etc., can the merchant actually see it and learn from it?
    From testing, attribution can be fragile. UTM on the original product/storefront landing page can survive into the order; adding attribution only at checkout often does not. So merchants should not assume all AI-driven sales will be obvious in analytics unless the signal is captured early.

The interesting part for me is loyalty. If an AI buyer is comparing 5 similar products, the final recommendation may not be based only on price. It may also consider “this shopper already has points/rewards with Store A” or “buying here earns a reward toward the next order.”

So I agree with the comments above: structured catalog data is the foundation. But I think the next layer is structured commerce value: rewards, discounts, policies, fulfillment, and customer-specific context that AI agents can safely query and explain.

Near term, I’d recommend merchants test three things:

  • Ask ChatGPT/Claude/Gemini what your store sells and see if it gets it right.
  • Check whether your product pages clearly answer “who is this for?” and “why this over alternatives?”
  • Start tracking whether AI-referred visits/orders are showing up correctly, because the attribution path is still early and inconsistent.