AI assistants surface products across physical and digital channels; can erode consideration when data incomplete
AI assistants surface products across physical and digital channels; can erode consideration when data incomplete The invisible shelf: Winning a place on the AI-curated shortlist AI is becoming a new gatekeeper of discovery. That creates what we call the invisible shelf—a new layer of product visibility that companies cannot fully see or control. AI assistants assemble recommendations from the information available to them, influencing which offers surface for a buyer and which do not enter consideration. Companies have spent years competing for visibility across channels, both physical and digital. Those still matter. The way companies compete for revenue, though, has fundamentally changed. Increasingly, capturing revenue will mean making what companies sell easy for AI to understand, evaluate and recommend so it can earn a place on the invisible shelf. Consumers already use AI to research products, compare options, interpret reviews and find deals. An IBM Institute for Business Value study found that 45% turn to AI for help during their buying journeys, including 41% who use it to research products. McKinsey has separately found that half of surveyed US consumers intentionally use AI-powered search. The customer still makes the final decision, but AI can influence which products enter consideration before a company sees a visit, query or lead. The invisible shelf is not a fixed catalog. The product shortlist that an AI assistant generates can change with the assistant, the wording of the question, the customer’s location and the information available at that moment. Companies can see where their products appear on their own sites and marketplaces, but they usually cannot see why an offer was absent from an AI-generated shortlist. Consider a customer looking for a laptop under USD 1,500 that is light enough for travel, compatible with several devices, covered for repairs across Europe and deliverable by Friday. A suitable product might be left out because its compatibility data is not structured for AI to find and use. Inconsistent service terms or an unconfirmed delivery date can create the same problem. No company can control every result. AI assistants use different sources, interpret requests differently and sometimes overlook relevant products. However, companies can reduce avoidable uncertainty by making the facts that influence choice clear, current, consistent and available in forms that digital systems can use. The commercial effect might be difficult to detect. If a product is omitted before the customer reaches a company channel, there might be no impression, site visit, internal search, abandoned cart or failed checkout to analyze. Not every omission represents a lost sale, but repeated absence from relevant shortlists erodes consideration without showing the company where or why it happened. Most attribution models begin after a customer touchpoint, so they cannot explain an opportunity that disappeared before one occurred. An AI assistant might find a product page and still lack the information needed to judge whether the offer fits the request. Claims such as “premium quality,” “fast delivery” or “ideal for families” communicate a position, but they do not resolve a specific need. Compatibility with named devices, repair coverage in particular markets, delivery to a defined postcode and certification against a recognized standard provide facts that can be applied to the decision. Those facts must agree wherever they appear. Structured product data can make information easier for digital systems to interpret, but it cannot correct stale inventory, conflicting terms or unsupported claims. A product record might look complete and still fail to answer the customer’s actual question. This shift brings operational performance into customer acquisition. Availability, promotion eligibility, delivery timing, installation coverage, warranty terms, service capacity and return rules were once treated mainly as concerns that followed customer choice. On the invisible shelf, they can influence whether a product is considered at all. The ability to keep the promise can now affect whether the offer is included. Proof matters for the same reason. A sustainability claim is more useful when its certification and scope are clear. An accessibility feature is easier to assess when its operation and limits are precise. A performance claim is stronger when the supporting test can be traced. Certifications, test results and policy details once served the legal, compliance or technical needs; they now also help determine whether a claim can be used in a comparison. In most companies, the commercial promise is divided across teams and systems. Product attributes, supporting evidence, price and eligibility rules, inventory, service terms and delivery commitments often have different owners. Customers see none of those divisions. They see one offer and expect every part of it to agree. A product page might promise next-day delivery while the order system cannot support it in the customer’s location. A warranty can appear broad while the detailed policy excludes the intended use. Compatibility might be documented but absent from the product data available to external systems. AI does not create these gaps, but it can expose their commercial consequences before the company knows that a customer is considering the product. Addressing the issue requires more than adding content. Companies need to identify the facts that most influence choice, establish an authoritative source for each one, assign clear ownership and make the information available in forms external systems can use. They also need to test realistic combinations of customer needs. Testing is important because an offer can look complete on paper, but fall short when location, timing, compatibility, eligibility and service requirements are considered together. Search rankings, traffic and conversion will remain important, but they cannot reveal every product that failed to enter consideration. Therefore, companies need to look upstream, at whether product data, evidence, policy and operations describe the same offer. When they do not, a company loses the chance to compete before it even knows that a customer was in the market. The goal is not to control every AI-generated recommendation. It’s to reduce the reasons a relevant offer might be excluded. That starts with identifying the facts that influence customer choice, establishing authoritative sources for them, keeping those facts current across systems and testing whether AI can use them in realistic buying scenarios. As AI becomes a more active intermediary between companies and customers, this discipline can turn the invisible shelf into a new arena for competition. Companies that make their offers easier to understand, verify and fulfill will be better positioned to earn consideration, wherever the customer journey begins. Explore how IBM Consulting® can help connect data and systems and orchestrate AI-powered workflows across the front office. This way, companies can create more intelligent customer experiences and make offers easier to understand, evaluate and act on across AI-enabled customer journeys.