.
Have OpenAI and Shopify Made a Mockery
of the Ethics of AI Search?
5th October 2026
The Shopify–OpenAI relationship may make AI shopping considerably better. But it raises a question the AI industry cannot avoid: if access to the information pipeline isn't equal, how meaningful is neutral selection — and how much does that cost AI search in trust and authenticity?
There is something genuinely impressive about what is happening between Shopify and OpenAI. Millions of products can potentially become available to people shopping conversationally through AI, with structured information about those products supplied through infrastructure designed specifically for the job.
For shoppers, that could make product discovery considerably better. For Shopify merchants, it removes some of the technical friction involved in making products available to an AI system. For OpenAI, it provides access to an enormous body of structured and regularly maintained commercial information.
There is nothing inherently wrong with any of that.
But there is an uncomfortable question sitting underneath it: what constitutes equality in AI-assisted discovery?
If an AI system selects products independently once information reaches it, is that enough? Or should businesses also have a reasonably comparable opportunity to get their information into the environment from which those selections are made?
That distinction matters because neutral selection and equal access are not necessarily the same thing.
What has actually been announced?
Shopify describes Shopify Catalog as “a structured source of product information gathered from eligible products that are sold by stores on Shopify.” It brings together product information that can then be made available to channels using the catalogue.
Source: https://help.shopify.com/en/manual/shopify-catalog
OpenAI has similarly explained that Shopify product data is integrated into ChatGPT through Shopify Catalog, helping products appear more accurately and completely in relevant conversations, with no additional work required from individual Shopify merchants.
Source: https://openai.com/index/powering-product-discovery-in-chatgpt/
Shopify's own documentation is even more explicit about the practical consequence. It describes ChatGPT as a “discovery-focused referrer platform”, says the channel is “active by default for eligible stores”, and tells eligible merchants: “You don't need to do anything for your products to be discovered by ChatGPT through Shopify Catalog.”
Source: https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/chatgpt
That is significant. An eligible Shopify merchant doesn't necessarily have to construct a bespoke OpenAI product feed, establish its own integration or independently solve every part of the product-data ingestion problem. Shopify has built infrastructure that can do a substantial part of that work on the merchant's behalf.
But the small print matters
It would be easy to turn this into a story about Shopify merchants buying their way into ChatGPT. Based on the evidence currently available, that would be wrong.
Shopify explicitly states:
“Being included in Shopify Catalog doesn't guarantee that a product will appear in a specific AI answer, be ranked in a specific position, or be displayed by every channel that's connected to Shopify Catalog.”
Source: https://help.shopify.com/en/manual/shopify-catalog
OpenAI also says product results are selected independently and that its partnerships do not influence which products are selected. There is therefore no evidence that Shopify has bought preferential rankings, recommendations or citations from OpenAI, and Citewell is not alleging that it has.
That qualification is important because there are two different questions here. The first is whether Shopify products receive preferential treatment when products are selected. We have no evidence that they do. The second is whether Shopify has reduced the friction involved in getting eligible product information into the environment from which those selections can be made.
The answer to the second question appears much clearer.
That is not preferential ranking. It is reduced distribution friction.
A merchant could have done almost no deliberate AEO or GEO work
This is where the arrangement becomes particularly interesting from an AEO and GEO perspective.
A Shopify merchant could theoretically have a mediocre wider website, heavily templated supporting pages, generic product descriptions, little original expertise and almost no deliberate AEO or GEO strategy, while still having its eligible catalogue information supplied through a clean structured pipeline.
That doesn't mean ChatGPT will recommend it. Poor product information can still count against a merchant, and both Shopify and OpenAI make clear that inclusion does not guarantee placement. The quality and completeness of the information still matter.
But the merchant has nevertheless cleared one significant technical hurdle without independently solving the discovery and ingestion problem.
For years, businesses have been told that technical accessibility, structured data, crawlability, information architecture and machine readability matter. They still do. But arrangements like this demonstrate that sufficiently large platforms can increasingly solve parts of that problem centrally for the businesses operating within their ecosystem.
The open web has to make itself discoverable. A sufficiently powerful platform can negotiate the plumbing.
Shopify is becoming more than a storefront provider
There is another important development here that makes this bigger than a discussion about Shopify merchants.
Shopify now offers an Agentic plan intended, in Shopify's words, “for merchants that don't already use Shopify as their ecommerce platform.”
That changes the competitive picture considerably. Shopify does not necessarily need a merchant to move its entire storefront onto Shopify to become involved in that merchant's AI-commerce distribution.
A business could theoretically retain another ecommerce platform while using Shopify infrastructure to participate in agentic commerce channels. Shopify is therefore positioning itself not simply as somewhere businesses build online shops, but increasingly as part of the infrastructure through which product information reaches AI-mediated shopping environments.
From Shopify's perspective, that makes considerable strategic sense. From the perspective of the wider ecommerce industry, it deserves close attention.
What happens to every other ecommerce platform?
Historically, ecommerce platforms competed on fairly familiar territory: storefront capability, checkout, flexibility, integrations, performance, cost, developer experience and SEO.
AI-mediated commerce introduces another consideration:
How easily can this platform make my products available to the AI systems through which customers increasingly discover them?
That question potentially changes purchasing decisions.
A merchant comparing Shopify with WooCommerce, Adobe Commerce, BigCommerce, a bespoke ecommerce engine or a smaller specialist platform may increasingly care about more than the quality of the storefront itself. It may also care about the platform's relationships with ChatGPT, Gemini, Copilot and whatever comes next.
That doesn't mean competing ecommerce platforms have been locked out. OpenAI is developing broader commerce infrastructure, including the Agentic Commerce Protocol, direct product feeds and third-party provider routes. Shopify is not the only way into AI commerce.
But the competitive landscape has nevertheless changed. An independent ecommerce platform may increasingly find itself competing not only against Shopify's software, but against Shopify's distribution relationships and infrastructure.
And Shopify's Agentic plan makes that particularly interesting. If a merchant moves to Shopify, Shopify wins the platform relationship. If the merchant remains elsewhere but uses Shopify as part of its AI-commerce distribution infrastructure, Shopify can potentially insert itself into that relationship anyway.
OpenAI does have another answer
This is where criticism of the arrangement needs to remain balanced.
OpenAI isn't saying that Shopify is the only route into ChatGPT commerce. Its Agentic Commerce Protocol is intended to provide a broader standard for connecting merchants and commerce systems with AI agents. OpenAI also supports direct product feeds and third-party providers.
Its developer documentation currently states:
“Onboarding product feeds in ChatGPT is currently available to approved partners.”
Source: https://developers.openai.com/commerce/guides/get-started
The words “currently” and “approved partners” matter.
OpenAI has indicated that broader merchant access is being developed. If genuinely accessible self-service ingestion becomes widely available, much of the access asymmetry discussed here could narrow considerably.
But it isn't unreasonable to examine the position as it exists today. Shopify's catalogue is already integrated. Individual eligible Shopify merchants don't have to establish their own OpenAI relationship. Other routes exist, but they can involve different processes, requirements or approval mechanisms.
That is why the important question isn't whether Shopify has bought ranking preference. There is no evidence it has.
The more interesting question is whether businesses currently have a reasonably equal opportunity to supply the information from which an AI system subsequently makes an independent selection.
Product data is not product knowledge
There is also a danger that Shopify merchants interpret all of this as meaning their AI visibility has been solved.
It hasn't.
A product catalogue can tell an AI that a jacket costs £89, comes in three colours, is available in several sizes and is currently in stock. It may contain images, identifiers, variants, descriptions and other structured attributes.
That doesn't necessarily tell the AI whether the jacket is suitable for Scottish winter rain, why its material was chosen, how its sizing compares with another brand, what customers frequently misunderstand about it, who should buy an alternative product, how it should be cared for or what years of experience have taught the retailer about the people who actually use it.
Those answers live somewhere else. They may be on the website, in buying guides, support conversations, manufacturer documentation, staff experience or simply inside the heads of people working for the business.
Product data and product knowledge are not the same thing.
Shopify can improve the route by which catalogue data reaches an AI system. It cannot manufacture knowledge the organisation has never expressed.
That distinction becomes increasingly important as AI shopping moves beyond finding products and starts helping people compare them, understand them and decide which one is appropriate.
Shopify merchants shouldn't celebrate too quickly
Being included in Shopify Catalog is therefore not the same as having a strong AI presence.
A merchant could have excellent access to one AI-commerce channel while still having an inadequate wider digital presence: weak product explanations, unanswered customer questions, missing specifications, generic manufacturer descriptions, poorly expressed expertise and very little evidence for AI systems discovering information through the wider web.
Different AI systems will not necessarily use identical sources, commercial relationships or retrieval mechanisms either. Building an entire AI-discovery strategy around one platform relationship therefore introduces another form of dependency.
Shopify's own data-privacy documentation provides an interesting glimpse of the distinction. It explains that when access to Shopify Catalog is blocked, AI channels may still crawl the merchant's online store, but the resulting information “might not always be as complete, accurate, or up-to-date.”
Source: https://help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/data-privacy
That is precisely why structured feeds are valuable. But it also demonstrates that there are now different routes by which machines can acquire information about the same business.
For merchants, the sensible response isn't to reject those routes. It is to avoid mistaking one particularly efficient route for the entirety of their digital presence.
If every Shopify merchant gets the same advantage, who actually wins?
There is another consequence worth considering.
Imagine two merchants selling broadly comparable products. Merchant A has spent years creating detailed buying guidance, accurate specifications, useful FAQs, original photography, strong technical implementation and genuine expertise around what it sells. Merchant B has done relatively little beyond maintaining basic catalogue information and generic product pages.
Shopify Catalog can potentially improve the machine accessibility of product information for both.
That does not mean Merchant B will outrank Merchant A. It does not mean the work done by Merchant A suddenly becomes worthless. In fact, richer information and genuine expertise may become even more important once basic catalogue distribution becomes commonplace.
What has changed is that one dimension of competitive differentiation — getting basic structured product information efficiently into participating AI-commerce systems — can increasingly be supplied by the platform itself.
Once that infrastructure becomes normal, the competitive question moves further up the knowledge chain. If everybody can provide the machine with a product catalogue, businesses increasingly need to ask what information they possess that makes their products genuinely more understandable, useful or appropriate.
If AI shopping systems become excellent at ingesting catalogues but poor at recognising genuine expertise, we will have solved the wrong problem.
Is AI search citation now up for sale?
This is where the argument needs particular care.
There is no evidence that Shopify has purchased ChatGPT citations, rankings or recommendations. OpenAI explicitly says partnerships do not influence product selection, and nothing we have found justifies claiming otherwise.
But commerce integrations force us to confront a broader question: can sufficient commercial scale negotiate a better route into the information environment from which recommendations are eventually selected?
That is not the same allegation.
Search engines have always had to decide what information enters their index and how that information is evaluated. AI systems introduce new ingestion mechanisms alongside traditional crawling: product feeds, APIs, licensed datasets, partnerships, protocols and structured commercial catalogues.
Those mechanisms can make AI systems substantially better. Fresh inventory and accurate prices are obvious examples where a maintained product feed may be superior to periodically crawling a web page.
But the architecture matters because who gets to supply the candidate set matters.
If some organisations can supply information automatically through deeply integrated infrastructure while others have to establish different routes into the same information environment, neutral ranking alone doesn't entirely answer the fairness question.
How much does this cost AI search in trust and authenticity?
Traditional advertising makes its commercial nature reasonably obvious. A sponsored result can be labelled as sponsored. A banner advert looks like advertising.
Conversational AI is different.
When an AI assistant tells somebody that a particular product is worth considering, the recommendation feels like the result of reasoning. The user may reasonably assume that the system has considered the available evidence and selected something because it appears appropriate.
That makes the provenance of the candidate information unusually important.
If some product catalogues arrive automatically through deeply integrated commercial infrastructure while other merchants currently have to apply for direct-feed access, use other providers or rely more heavily upon wider discovery mechanisms, businesses and users are entitled to ask how much that difference matters.
OpenAI may have a perfectly reasonable answer. Its development of ACP, multiple feed mechanisms and broader merchant access suggests it recognises that wide participation matters.
But as AI systems increasingly mediate commercial decisions, the burden on those systems becomes greater. It may no longer be enough to demonstrate that selection is neutral. They may increasingly need to demonstrate that access is reasonably open.
Because once users begin wondering whether an AI recommended something because it was the best available answer or because that information had the easiest route into the system, something valuable begins to erode: trust.
And ultimately, authenticity.
Regulators may eventually ask the same question
This issue also extends beyond Shopify and OpenAI.
European regulation and competition policy are increasingly concerned with digital gatekeepers, interoperability, self-preferencing, access to data and the ability of businesses to compete fairly within large digital ecosystems. That does not mean the Shopify–OpenAI arrangement breaches European law, and Citewell is not suggesting that it does.
But AI commerce creates precisely the sort of new market structure regulators will eventually have to understand.
If a handful of large commerce platforms become the easiest suppliers of structured commercial information to the dominant AI assistants, regulators may ask whether businesses outside those ecosystems have a materially comparable opportunity to participate.
Europe may therefore become one of the first serious tests of what fair access means when AI assistants become intermediaries between businesses and consumers.
The answer shouldn't necessarily be to prevent integrations. Structured commerce data is useful. The more difficult challenge is ensuring that useful infrastructure doesn't quietly become another generation of gatekeeping.
Efficiency is not equality
None of this makes Shopify's strategy irrational. Quite the opposite.
Shopify has accumulated enormous product-data scale and is using that scale to make itself valuable to the systems through which commerce may increasingly happen. OpenAI needs accurate, structured and current product information. Shopify can provide a substantial amount of it.
From an engineering and commercial perspective, the relationship makes sense.
The ethical question is whether efficiency for the largest participants ultimately produces a sufficiently open environment for everybody else.
That distinction matters far beyond ecommerce. AI systems are increasingly becoming interfaces through which people discover businesses, compare services, learn about products and decide what information to trust. The mechanisms controlling what enters those systems may therefore become almost as consequential as the algorithms deciding what comes back out.
AI search cannot claim openness merely because its final selection process is neutral if meaningful differences exist in the routes by which information becomes available for selection.
That doesn't prove the system is unfair.
It means the question deserves an answer.
Where Citewell stands
Citewell exists to help organisations build an ethical online digital presence based on what they genuinely know, what they can substantiate and what they are prepared to put into the public information environment.
We believe AI-assisted discovery should reward useful information, genuine knowledge, clear evidence and businesses that make themselves understandable. Product feeds have a legitimate role in that. Structured data has a legitimate role. Commercial integrations have a legitimate role too.
But none should become a substitute for an open and reasonably level opportunity to be discovered.
Our concern is therefore not that Shopify has made its merchants' product information easier for AI systems to consume. That is useful infrastructure, and other commerce platforms will undoubtedly develop their own responses. The more important question is whether the emerging AI information environment remains sufficiently open that organisations can compete through the quality of their knowledge and evidence rather than primarily through the strength of the distribution ecosystem to which they belong.
Citewell will continue helping businesses make their knowledge clearer, more useful, more attributable and more accessible across the wider digital environment rather than encouraging them to build their entire presence around whichever platform happens to have the best relationship with today's AI systems.
Platforms change. Partnerships change. Search systems change. The information environment will change again.
Your knowledge should remain yours.
So, have OpenAI and Shopify made a mockery of the ethics of AI search?
Not in the simplistic sense that Shopify has bought recommendations or that ChatGPT is secretly ranking Shopify merchants because of a commercial agreement. We have found no evidence for either proposition, and the available documentation explicitly points in the other direction.
But the relationship exposes a problem the AI industry will have to confront.
There is a difference between saying that everybody who reaches the starting line will be judged fairly and ensuring that everybody has a reasonably fair opportunity to reach that starting line.
Shopify has built an efficient route for enormous quantities of product information to reach emerging AI-commerce systems. OpenAI, to its credit, is also developing broader routes through ACP, direct feeds, third-party providers and planned wider merchant access.
How genuinely open those routes become may determine whether today's concern eventually disappears or grows into something much larger.
AI search has an opportunity to improve on some of the gatekeeping structures that characterised earlier generations of the internet. It would be a mistake to recreate those structures simply because the new gatekeepers speak in natural language.
The ethical answer shouldn't be to ensure that every business can eventually find the right gatekeeper.
It should be to reduce how much businesses need a gatekeeper in the first place.