Introduction
Imagine asking an AI assistant a simple question: “Find me the best noise-cancelling headphones under US$300.”
The system does not show you hundreds of products. It compares specifications, reads reviews, checks prices and filters the market into a small set of options. It may ask whether you care more about sound quality, comfort or battery life. In some cases, it can guide you from the first question to the final transaction without requiring you to visit multiple retail websites.
Before you see a product, the AI has already decided that it deserves to be seen.
For decades, marketing operated on a familiar assumption: people made purchasing decisions, and brands competed for their attention. Companies invested in advertising, packaging, celebrity endorsements, retail design and digital campaigns to reach the right consumer with the right message.
Artificial intelligence is adding a new layer to that relationship.
Brands are no longer communicating only with people. They are also communicating with systems that interpret product information, compare alternatives and increasingly act on behalf of shoppers. These systems can influence which products enter the consumer's field of consideration and which disappear before the consumer knows they exist.
AI is not the customer in a literal sense. It does not experience desire, spend its own money or live with the consequences of a poor purchase. The consumer remains the principal—the person who defines the need, provides the budget and carries the risk.
But AI is becoming the delegated gatekeeper.
It increasingly determines which parts of the market the customer sees.
From Search to Delegated Choice
Traditional online shopping required consumers to perform much of the work themselves. Search engines presented links, e-commerce platforms displayed product listings and shoppers moved between websites to compare prices, features and reviews.
Generative AI changes this process because it does not simply retrieve information. It interprets information and turns it into an answer.
Instead of searching for “best running shoes 2026” and reading ten articles, a consumer can describe their budget, running habits, injury history and preferred style. The AI can then return a much shorter list of products, explain the trade-offs and refine the recommendation through conversation.
This movement from search to delegated choice became more visible in 2025. OpenAI introduced Shopping Research, an experience that examines product information across the internet and produces personalised buying guides. It also introduced Instant Checkout for selected merchants, allowing some users to move from product discovery to purchase within ChatGPT. In 2026, ChatGPT expanded its product-discovery experience with visual browsing, side-by-side comparisons and more detailed information on prices, reviews and features (OpenAI, 2025a, 2025b, 2026a).
Amazon's Rufus follows a similar logic. It draws on product listings, customer reviews, community questions and information from across the web to answer questions, compare options and recommend products. Google is also developing product information designed specifically for conversational and AI-driven discovery.
These tools remain at different stages of development, and AI shopping is not yet universal. Capgemini Research Institute found that 25% of consumers had used generative-AI shopping tools in 2025, while a further 31% planned to use them in the future. The transition is still unfolding, but its direction is becoming clear.
The consumer journey is moving from active comparison towards assisted interpretation.
That shift creates a new form of market power. When an AI system turns thousands of products into five recommendations, it does more than save time. It controls the shortlist.
A brand excluded from that shortlist may never receive the opportunity to persuade the consumer.
Marketing Must Speak to Desire and Evidence
Traditional branding is built around meaning.
Luxury brands do not simply sell leather, fabric or metal. They sell taste, status, memory and belonging. Sports brands sell confidence and aspiration rather than shoes alone. Beauty brands connect products with identity, intimacy and personal transformation.
These meanings still matter. AI has not made emotional branding obsolete.
But emotional desire and algorithmic legibility are not the same thing.
A consumer can value a story because of how it feels. An AI system is more likely to act on that story when the claims behind it are accessible, consistent and supported by evidence.
AI shopping systems combine structured product information with unstructured material. They may examine specifications, prices, availability, customer reviews, product pages, community questions, editorial coverage and previous conversational context. They can process storytelling and emotional language, but they must also determine whether a product satisfies the consumer's stated conditions.
For marketers, this creates a two-part communication challenge.
The first language remains human. It uses design, narrative, experience and cultural meaning to make a product desirable.
The second language is designed for machine interpretation. It uses accurate product information, clear comparisons, consistent terminology, reliable sources and evidence that can be retrieved and verified.
The second language is not simply code.
It is evidence.
A beautifully written campaign cannot compensate for a product page with contradictory specifications. A strong brand story will struggle to travel through AI systems if the product name changes across platforms, prices are outdated, variants are unclear or important information is locked inside images that cannot be interpreted reliably.
The most competitive brands will therefore combine emotional distinctiveness with informational clarity. They will make people want the product while making it easy for AI systems to understand why the product fits a particular need.
Trust Is Becoming Machine-Readable
Consumers experience trust as a feeling. AI systems infer it through signals.
These signals can include accurate specifications, consistent product descriptions, credible third-party sources, transparent pricing, clear return policies, verified reviews and reliable information across different platforms.
This does not mean that an algorithm genuinely trusts a brand. It means that certain features associated with credibility can be represented, compared and retrieved.
Trust, in this limited sense, is becoming machine-readable.
Google's Merchant Center offers a practical example. Its structured product data allows platforms to understand information such as price, availability, shipping and product variants. Google has also introduced “conversational attributes” that allow merchants to submit product questions and answers, document links, related products and variant information. These attributes are explicitly designed to help AI systems and conversational agents understand product nuances and surface them across AI-driven experiences.
This is more than a technical change. It affects how marketing, e-commerce, customer service and reputation management work together.
Every product description contributes to the brand's machine-readable identity. Every customer review becomes potential evidence. Every inconsistency between a website, a retailer listing and a customer-service response creates uncertainty that an AI system may reproduce.
The growing interest in Generative Engine Optimisation, or GEO, reflects this shift. Aggarwal et al. (2024) define GEO as a framework for improving the visibility of content within generative-engine responses. Their experiments found that techniques such as adding credible citations, relevant statistics and clearer information could improve source visibility, although the effects differed across topics and platforms.
GEO should not be treated as a guaranteed formula. AI systems remain proprietary, fast-changing and partly opaque. A method that improves visibility in one engine may have little effect in another.
Still, the underlying principle is important: brands must consider whether intelligent systems can locate, interpret and justify their claims.
Visibility is no longer only about ranking first on a search page.
It is also about becoming part of the answer.
Platforms Still Shape What AI Recommends
It would be misleading to suggest that AI shopping creates a neutral marketplace where the most trustworthy product automatically wins.
AI systems operate inside commercial platforms. Those platforms have their own business models, merchant relationships, data partnerships and technical requirements. Their recommendations may be shaped by source availability, product coverage, popularity signals and the information they are able to access.
Advertising has not disappeared either.
In 2026, OpenAI began testing advertising in ChatGPT. The company states that advertisements are clearly labelled, separated from organic responses and do not influence the answers ChatGPT provides. Even with this separation, the development demonstrates that AI interfaces can contain both organic recommendations and paid product discovery (OpenAI, 2026b).
Brands may therefore compete in three connected markets.
The first is organic recommendation: whether the system independently considers the product relevant.
The second is paid visibility: whether the brand appears through a clearly identified advertising placement.
The third is platform accessibility: whether the product is included in the feeds, merchant systems and technical infrastructure that allow the AI experience to understand and display it.
This makes algorithmic visibility partly a marketing challenge and partly a question of platform governance.
The future will not be defined by “trust instead of advertising.” It will be defined by the relationship between trust, commercial visibility and the rules established by AI platforms.
What Brands Need to Build Now
The rise of AI shopping does not require brands to abandon traditional marketing. It requires them to expand what they consider a brand asset.
The first priority is a reliable product-information layer. Product names, prices, availability, specifications, variants, compatibility information, shipping terms and return policies should be accurate and consistent across the brand website, retail platforms and merchant feeds. Structured data should support what consumers can already see rather than contradict it.
The second priority is an evidence architecture. Brands need more than promotional claims. They need product documentation, original research, independent reviews, credible media coverage, customer questions and clear explanations of how products differ. These sources give AI systems material they can use when explaining why a product fits a consumer's request.
The third priority is AI visibility auditing. Companies should regularly ask major AI platforms the questions real consumers are likely to ask:
· Is the brand mentioned?
· Is the product recommended?
· Which competitors appear first?
· Are the specifications and prices accurate?
· Which sources are cited?
· Does the answer reflect the brand’s intended position?
· Does the recommendation change across languages, platforms or different versions of the same question?
This should not be a one-time exercise. AI results can change as models, sources, product feeds and consumer contexts change.
The fourth priority is reputation quality. Reviews should not be treated only as social proof placed at the end of the purchasing journey. They increasingly form part of the evidence that future recommendation systems may summarise. Customer service, delivery and after-sales support therefore contribute not only to retention but also to future product discovery.
Marketing no longer ends when the transaction is completed.
Every transaction can become evidence for the next recommendation.
The fifth priority remains human distinctiveness. If several products satisfy the same technical requirements, the final decision may still depend on recognition, identity, aesthetics, experience and emotional connection. Machine-readable information can help a brand enter the shortlist, but it cannot guarantee that the consumer will care.
Brands still need a reason to be chosen after they have been found.
The Consumer Remains the Principal, but Not Always the Final Click
There are limits to what AI can determine.
An AI assistant can compare the ratings of perfumes, but it cannot personally experience how a fragrance develops on someone's skin. It can identify a technically suitable family car, but it does not feel the driving position or understand the symbolic meaning the vehicle may carry for its owner.
These limits are especially important in fashion, beauty, luxury, travel and other categories where purchasing is tied to identity and embodied experience.
Yet it is also too simple to say that the human will always make every final decision.
For emotionally significant, expensive or unfamiliar purchases, consumers are likely to remain closely involved. They may want to touch, test, smell, drive or experience the product before buying it.
For routine, low-risk or repeat purchases, more decision-making may be delegated. A consumer might allow an AI agent to reorder household products, purchase an item when it falls below a target price or select between products that meet previously defined conditions.
In these situations, the human still establishes the goal and boundaries, but the agent may select the product and execute the transaction.
The more authority consumers delegate, the more important transparency becomes. Capgemini found that 71% of consumers were concerned about a lack of clarity around how generative AI collects and uses personal data. Consumers may appreciate personalisation while remaining uncomfortable with the amount of information required to produce it.
The future of AI shopping will therefore depend not only on convenience, but also on whether consumers understand why a product was recommended, which information shaped the decision and whether commercial relationships influenced what they saw.
The Customer Has Not Disappeared. The Gatekeeper Has Changed
When AI shops for you, the real customer is still the person whose need is being served, whose money is being spent and whose life is affected by the purchase.
But that person is no longer the only audience a brand must persuade.
AI systems increasingly determine which products are visible, comparable and easy to justify. Commerce platforms determine which information those systems can access and how organic recommendations, merchant participation and advertising coexist.
The new purchasing journey is therefore not simply:
Brand → Consumer → Purchase
It is increasingly:
Brand → AI and Commerce Platform → Consumer → Purchase
The winning brands will not choose between emotional branding and AI optimisation. They will build both.
They will create products and stories that people genuinely value. They will also create accurate, transparent and machine-readable evidence that intelligent systems can confidently interpret.
Consumers remain the destination.
But AI is becoming the gatekeeper that determines which brands reach them first.
References
Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., & Deshpande, A. (2024). GEO: Generative engine optimization. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.
https://doi.org/10.1145/3637528.3671900
Amazon. (2024). Amazon announces Rufus, a new generative AI-powered conversational shopping experience.
https://www.aboutamazon.com/news/retail/amazon-rufus
Capgemini Research Institute. (2026). What matters to today’s consumer 2026.
Google Merchant Center. (n.d.). About structured data markup for Merchant Center.
https://support.google.com/merchants/answer/6069143
Google Merchant Center. (n.d.). How to use conversational attributes.
https://support.google.com/merchants/answer/17085370
OpenAI. (2025a). Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol.
https://openai.com/index/buy-it-in-chatgpt/
OpenAI. (2025b). Introducing shopping research in ChatGPT.
https://openai.com/index/chatgpt-shopping-research/
OpenAI. (2026a). Powering product discovery in ChatGPT.
https://openai.com/index/powering-product-discovery-in-chatgpt/
OpenAI. (2026b). Testing ads in ChatGPT.