No matter how renowned your brand is, the system might not "see" it at all.

When consumers start asking AI "what to buy," the rules of brand competition have fundamentally shifted. Research shows that household names like Nike and Disney are being outperformed by niche brands in AI recommendations—because AI does not prioritize the "most famous" brands, but the "most interpretable" ones. Whether your brand has clearly defined attributes and verifiable evidence directly determines if it will be included in AI-generated recommendations.

When we asked mainstream AI systems—ChatGPT, Claude, and Gemini—to recommend running shoes, the relatively niche brand Brooks consistently appeared across results. Meanwhile, Nike, the world's largest sportswear brand, showed up far less frequently. This pattern reflects a fundamental shift: as AI systems become the primary medium for product discovery, how brands compete has changed entirely.

Brooks did not build its brand around broad lifestyle narratives, but focused instead on technical performance and meeting the specific needs of dedicated runners. Under the leadership of CEO Jim Weber, the company narrowed its scope, exited adjacent product categories, and invested heavily in biomechanical research and product engineering. Technologies like GuideRails and DNA LOFT cushioning were developed to solve clearly defined pain points. Equally critically, Brooks cultivated an ecosystem of coaches, clinicians, and specialty retailers who could break down these technical solutions using precise, specific terminology. In other words, Brooks built a brand that is "algorithmic-readable."

AI systems favor brands that can be broken down into concrete attributes and factual evidence—brands whose value proposition can be clearly and explicitly articulated when a user asks about their advantages. In a study covering 15 retail categories—including laptops, pet food, and credit cards—we tested GPT-4o, Claude, and Gemini using identical prompts. Across all tests, brands were mentioned more than 1,000 times, spanning 716 distinct brands. Conducted by Georgetown University's McDonough School of Business and the University of Virginia's Darden School of Business, this research revealed a consistent pattern: brands are no longer primarily competing for human attention.

AI tools are rapidly emerging as the new frontier entry point for product discovery. Unlike search engines and traditional media that present brands based on visibility or storytelling, AI systems are designed to help consumers make active choices. In the AI landscape, brands compete to earn a spot on the model's shortlist of candidates during recommendation generation—a goal most existing brands were never built to achieve.

AI Recommends What It Can Interpret

Our research uncovered four patterns that collectively explain why so many established brands are losing ground in AI-driven product discovery—and why interpretability is the key to reversing this trend.

First, what marketers refer to as "AI visibility" is far more fragmented than it appears. Of the 716 brands identified in our study, only 8.4% appeared consistently across all three platforms: ChatGPT, Claude, and Gemini. Most brands showed up on just one single platform. A brand that seems prominent in one system may be completely absent from another.

Brands may still be pouring resources into boosting their public visibility, but that is not what determines whether AI systems recommend them. The critical factor is whether the model can recognize your brand as a reliable solution to a specific user problem. When a brand's attributes and supporting evidence are clearly organized, different systems are far more likely to independently identify it as a valid recommendation. Without that structure, a brand's presence will be inconsistent, intermittent, or disappear entirely.

Second, 55% of brands that appear across multiple platforms receive different positioning labels across different systems. A brand categorized as a "premium innovator" on one platform might be labeled an "affordable alternative" on another. This happens because AI systems do not directly import pre-packaged brand messaging—they infer your positioning from third-party information they can access. The brand profile a model constructs is based on verifiable attributes and supporting evidence, not the narrative your marketing team wants to promote. Symbolic branding is nearly useless unless it translates into concrete, system-recognizable attributes.

This discrepancy is starkly visible in our data. Apple appears consistently across multiple platforms in the laptop and headphone categories, and Sony achieves near-perfect cross-platform consistency for headphones. Yet many of the world's most globally recognized brands are missing entirely. Disney, Starbucks, McDonald's, Netflix, IBM, and Intel do not appear in any of our query results. Even well-known brands that do appear are often represented by an interpretable sub-brand rather than the parent brand itself. Toyota shows up through specific models like the RAV4 and Highlander; Coca-Cola and Pepsi appear through their zero-sugar product lines. In these cases, AI makes judgments based on specific product attributes, not the symbolic value of the parent brand.

Third, the way queries are phrased defines the competitive landscape. Exploratory queries generate 95% more brand mentions than narrowly targeted queries; and only around 11% of brands appear in the results of both query types. AI assistants build recommendations based on how consumers describe their needs. A user asking for "running shoes" gets one set of candidates; a user asking for "running shoes for people with knee pain" or "stability running shoes for overpronation" gets a completely different shortlist.

There is an even deeper influencing factor here: brands can shape the exact vocabulary consumers use to describe their problems. Brooks spent 20 years teaching runners to name their specific issues: overpronation, gait deviation, and weight-bearing stability. These terms spread through coaching circles, specialty stores, and running media. Brands that invest in this kind of "problem literacy" create a favorable query pathway for themselves long before any AI generates a recommendation.

The fourth finding completes the full picture: 78.7% of all brand mentions carry positive sentiment, and this percentage is remarkably consistent across all three platforms. Once a brand is included in a recommendation, it almost always receives favorable commentary.

This directly reflects how AI systems operate. They first determine which brands qualify as valid answers to a user's question, and only then generate positive framing. Traditional media, by contrast, competed for exposure volume and favorable public reputation.

The Real Competitive Bottleneck

Is Being Selected, Not Being Praised

The truly critical strategic question is not "how do we get AI to say nice things about us?" but "how do we get our brand to appear in AI responses in the first place?" The answer is interpretability.

A brand will appear in AI answers when the model can construct a clear logical chain, connecting the user's specific context to product features, and then to the brand that satisfies that need. AI recommendations do not start with the brand and its marketing promises. They start with the user's conditions, framed by the query, and work forward step by step: User Context → Product Need → Brand That Fulfills the Need.

Interpretability depends on three core elements:

Entity Clarity: The brand can be unambiguously identified across all information sources;

Structured Attributes: Product features have defined names that can be compared and measured;

Corroborating Foundation: The brand's claimed advantages are supported by reliable, independent sources.

Brands with these characteristics are far more likely to be included in AI system recommendations, because their attributes and supporting evidence can be clearly linked to user needs.

At its core, competing for AI recommendations is fundamentally an information architecture problem that requires cross-functional collaboration. In most companies, brand messaging belongs to the marketing department, product specifications are managed by engineering and product teams, and third-party validation (reviews, expert assessments, clinical data) often lacks clear ownership. AI recommendations exact a high cost for this siloed approach. The brands most likely to win are those that establish cross-functional accountability around "how the brand is understood and retrieved as a solution."

This redefines what brand building should accomplish. Traditional brand building invests in storytelling, symbols, and emotional positioning to capture human attention and influence feelings and memory. AI systems rely on something entirely different: structured attributes, measurable product features, and verifiable facts that tie the brand to specific user problems. You are no longer marketing to human decision-makers—flattering narratives do not work, hard evidence does.

Three Ways to Improve Your AI Inclusion Rate

Marketing executives have traditionally measured their performance using brand strength metrics like market share (which brands consumers buy) and mind share (which brands consumers think of first). AI-driven product discovery demands a new metric: AI Inclusion Rate—the frequency with which a brand is retrieved as a valid candidate when it can actually solve a user's problem.

This is not the same as the "model share" described by Dubois, Dawson, and Jaiswal. Model share measures how many times a brand appears in AI-generated responses, focusing on exposure volume. AI Inclusion Rate, by contrast, measures fit—how reliably a brand is retrieved when its attributes match a user's query. This fit determines whether the brand makes it onto the final recommendation list.

When consumers turn to AI assistants—for "running shoes for knee pain," "the best laptops for video editing," or "credit cards with great travel rewards"—the system identifies the underlying need behind the query, then pulls out brands with matching attributes. Interpretability makes it easier for the model to connect the user's situation to the brand's attributes and supporting evidence, thus boosting the AI Inclusion Rate. Brands now compete not just to be remembered by consumers, but to be findable by the AI systems that help consumers make choices.

For marketers, improving the AI Inclusion Rate requires rethinking how brands communicate their value. Three approaches are particularly critical.

1. Replace Subjective Claims With Verifiable Specifications

AI systems struggle to process vague, ambiguous claims. Replacing "high quality" with "1000-cycle durability, ISO-certified" provides the model with actionable, useful information. To be retrievable, brands must express their value through named, comparable features that directly relate to specific user needs.

This requires translating positioning into concrete specifications. Performance metrics, design parameters, or clinically validated outcomes allow AI systems to connect user conditions to product capabilities. The more precise and well-supported a claim is, the more likely it is to be used in automated recommendations.

Brooks embodies this principle. Its products are described through measurable attributes—stability ratings, heel-to-toe drop, and biomechanical features designed to address specific running conditions. These attributes make it far easier to link queries like "running shoes for knee pain" to the brand's solutions.

Sony and Apple reflect the same logic across different categories. Sony's products are defined by technical specifications such as noise cancellation performance and sensor capabilities, often backed by independent benchmark tests. Apple products similarly rest on measurable performance metrics, from processor benchmarks to battery life.

What these brands share is not just reputation and scale, but interpretability—the ability to frame their product features in terms that can be evaluated, compared, and used in the AI's reasoning process.

2. Cultivate Independent, High-Authority Third-Party Validation

Being included in AI responses depends on the ecosystem of reviews, expert assessments, and research surrounding your brand. Brooks built partnerships with specialty running stores, coaches, podiatrists, and clinicians—people who need to explain why a specific shoe helps a specific runner. AI systems recommend Brooks in part because the company spent 20 years making itself easy to interpret. The AI Inclusion Rate is the long-term payoff of sustained investment in third-party credibility.

3. Shift Focus From Symbolic Appeal to Evidence Structure

Many well-known, trusted brands appear infrequently in AI-generated recommendations because the way they communicate their advantages cannot be leveraged by models. Traditional brand building prioritizes emotional positioning—lifestyle associations, brand origin stories, and vague signals of quality. These build consumer favorability, but do not translate into structured attributes and factual evidence that AI systems can process.

As a result, brand recognition does not automatically translate to AI retrieval. Closing this gap requires a strategic shift: reduce reliance on symbolic claims, and strengthen the evidence base that connects your brand to specific user problems.

In an AI-driven market, brand strength depends not just on how a brand is known or perceived, but on how easily it can be identified as the answer to a specific problem.

Start With a Simple Diagnostic

Every brand needs to ask itself one strategic question: Where does my brand currently stand, and how can I make it more "interpretable"? You can begin with a straightforward diagnostic process.

First, use the keywords your customers actually use—whether category terms or specific problem descriptions—to query the major AI platforms: ChatGPT, Claude, and Gemini. See which brands appear, how your brand is described, and whether that description is consistent across different platforms. Our research found that 55% of brands are described differently across systems—meaning AI is assigning your brand a label that may not align at all with the positioning you want to communicate.

Second, audit your brand's attribute structure. Can an average customer—or even an AI—name three measurable, comparable product features of your offerings that directly correspond to specific user needs? If the honest answer is "no," you have work to do. The goal is not to write a technical manual, but to ensure the attributes that make your brand distinct have clear, consistent names that are used uniformly everywhere your brand appears.

Third, map out your third-party evidence base. Which independent voices—review organizations, experts, clinicians, specialized media—are using your key attributes to describe your products? Where are the gaps? The brands that perform best in AI recommendations are not the ones with the largest ad budgets, but those with the most consistent, credible external validation. This kind of corroboration is built over time and cannot be bought with advertising spend.

Pay close attention to the exact words your customers use to describe their problems—and actively shape that vocabulary. Invest in "problem literacy": give the specific problems your products solve clear, shareable names. These terms, spread through coaching networks, professional communities, and expert media, create a favorable query pathway for your brand long before any AI generates a recommendation.

The brands that gain the biggest advantage in AI-driven product discovery are not necessarily the ones spending the most on AI marketing today. They are the brands that spent years incrementally building a corroboration ecosystem that makes them easy for AI to find and understand.

AI-driven product discovery is not just another new media channel to optimize. It demands a structural shift in how brands compete—a transition that rewards a completely different set of capabilities. The mass media era competed for reach. The search era competed for relevance. The social media era competed for engagement. The AI assistant era competes for interpretability—the depth of attributes and factual evidence your brand provides that allows AI to generate compelling, logical recommendations on your behalf.

The brands that win in this environment will not be the most famous ones, but the ones that are most understandable to the reasoning systems that start from a user's problem and work backward to a solution. In the AI landscape, the chance to be shortlisted is the real competitive bottleneck. Once a brand makes it onto the candidate list, AI will almost always frame it positively. The real competition happens much earlier, determined by how your brand's attributes and supporting evidence are structured.

Brooks did not build an interpretable brand for AI. It built it for human experts who needed to explain to real runners why a specific shoe was the right choice. It turns out these two goals are identical. The brands that will win over the next decade are the ones that intentionally make that same choice.

John Gale, Luca Cian, Luc Wathieu | Article

John Gale is a consultant and adjunct professor at Georgetown University's McDonough School of Business. Luca Cian is a professor of marketing at the University of Virginia's Darden School of Business. Luc Wathieu is a professor of marketing at Georgetown University.

This article originates from the WeChat Official Account "Harvard Business Review" (ID: hbrchinese), written by HBR-China, edited by Zhou Qiang, and published with authorization from 36Kr.