Why brands need to be rethought in the age of AI.

For a long time, we have designed brand communication mainly from the perspective of human perception: what stands out? What gets remembered? What emotion does an encounter trigger? These questions remain valid. But they no longer describe every situation in which a brand meets people. Between sender and recipient there can now stand a system that summarizes, weighs and translates information into its own answer.

If I ask an AI assistant for a suitable solution, I may see neither a company's campaign nor its homepage. I receive an interpretation: this brand stands for something, fits a need, or is ranked against other offers. Which information is used for that depends on the particular system, for example training data, current search results, accessible sources or the context of the request. Not every answer is based on up-to-date research.

For me, this broadens the task of brand management. We must consider not only how a brand appears, but also what can be verifiably said about it. Are performance promises precise? Is it clear who an offer suits and who it does not? Do self-description, product information and independent statements align sufficiently? An attractive surface cannot replace answers to these questions.

A machine does not interpret the way a human does. It does not have lived experience, belonging or trust. Its answer can sound plausible and still be wrong. That is why I deliberately speak of an additional layer of mediation, not a new authority that objectively judges a brand's value.

Brands still need to be perceived. But increasingly they also need to be described in such a way that their interpretation is not left to chance.

How concretely this shift is already showing up in search can be seen in a study by the Pew Research Center. Pew analyzed the observed browsing behavior of 900 US adults and their Google searches in March 2025. The article, published in July 2025, describes differences between search result pages with and without an AI summary. This is a time- and geography-limited finding, not a general measurement of all AI use.

On search result pages with an AI summary, a click led to a classic search result in 8 percent of observed cases. On pages without such a summary, it was 15 percent. Users clicked a source link within the AI summary in 1 percent of visits to a search results page with an AI summary. These percentages refer to the observed search page visits and their click behavior respectively. They do not refer to brand awareness, market share, or the proportion of all people who use AI.

Nor do these figures prove that AI summaries alone cause the lower click rate. Queries that trigger such a summary may differ in content and intent from other queries. The observation initially shows a relationship between the appearance of search results and behavior within this particular research context. From it, one can derive neither a guaranteed traffic loss for every company nor an unchanged pattern for other countries.

Still, I find the finding strategically important: an encounter with a brand can occur without a visit to its own channels resulting from it. People can receive information, rule out an offer, or make a preliminary choice while staying within another provider's interface. Reach on a brand's own pages captures only part of such moments.

I would therefore not simply optimize for the next click. I would first examine what questions potential customers ask and what they learn about the brand before experiencing it directly. This changes the perspective on touchpoints: not every relevant contact belongs to us, and not every relevant contact leaves a trace in our web analytics.

For me, a branding space is not just a surface we can design. It is a context in which a brand acquires meaning. Today that also includes a conversational answer, an AI-assisted product consultation, an automatically generated comparison, or an assistant's recommendation. Some of these spaces are shaped by companies themselves. Others lie outside their direct control.

Within these contexts, the concrete task facing people gains weight. The question may no longer be: which brand do I know? But rather: which solution fits my situation, my budget and my requirements? A brand must not only be mentioned there, but meaningfully placed. A vague promise like "innovative solutions for the future" provides little orientation for that. Clear use cases, understandable differences and verifiable performance limits provide considerably more.

This does not mean branding should be reduced to factual product features. Tone, stance, and the way a brand responds to uncertainty remain part of the experience. When a company offers its own assistant, new design questions arise: how does it explain things? When does it ask follow-up questions? How does it handle a request it cannot fulfill? Where does its remit end? Here, brand identity becomes observable behavior.

With third-party systems, the possibilities are more limited. We cannot order up any particular phrasing and cannot force a recommendation. But we can ensure that accessible information is not contradictory, outdated or empty of substance. I see this as a connection between brand strategy, editorial work, product communication and digital quality, not simply another advertising discipline.

The new space lies between question and answer. Brand management must understand what happens there, even if it cannot control every detail of that space.

To make this task tangible, I think of AI readability as three interconnected layers. This is my strategic working model, not a scientifically validated standard or a technical certification. It helps me consider human meaning, reliable knowledge and behavior at new touchpoints together, rather than splitting them into separate responsibilities.

The human layer describes what the brand stands for in people's eyes: its identity, its benefit, its stance, its distinctive voice and the experiences that support its promise. This is where it is decided whether a brand has anything relevant to say in the first place. Machine accessibility cannot compensate for a lack of meaning. I do not want to develop brands that are excellently readable and at the same time completely interchangeable.

The knowledge layer translates this foundation into robust information. This includes clear descriptions of the company and its offerings, consistent naming, current facts, traceable sources and understandable relationships between services, target groups and use cases. Technical accessibility and sensibly deployed structured data can support this. But they are no substitute for good content, and formally correct markup does not make a claim true.

The interaction layer concerns the actual encounter: how is a question understood? What information is helpful in which situation? When must an answer state a limit, ask a follow-up question, or hand over to a human? For a company's own AI applications, this layer can be designed and tested. For external assistants, it is more about observing how they represent the brand and fixing errors at the sources that can actually be influenced.

The limits of this model are explicitly part of it. Different systems draw on different information, answers can vary, and sources can be missing. Readability guarantees neither mention nor correct interpretation. Even Google notes, for its AI search features, that established SEO fundamentals remain relevant and that no special additional AI markup is required. For me, the conclusion is: no magic markup, but clarity, accessibility and verifiable substance.

Human, knowledge and interaction must work together. A brand does not become future-ready by trading its human side for machine order.

AI can accelerate drafts, generate variants and adapt content for different contexts. But that is exactly where a strategic danger lies: if production grows faster than clarity, we mainly scale vagueness. More text does not automatically make a brand more understandable. In an unfavorable case, it produces ever more phrasings for the same offer, contradictory statements, or a tone that sounds professional but no longer belongs to anyone.

Adobe's AI and Digital Trends Report 2026 looks, among other things, at companies' readiness for AI-supported customer experiences and agentic applications. I read such an industry report as guidance on expectations, priorities and organizational challenges. However, surveyed intentions or assessments are not the same as demonstrated implementation, nor proof that a particular tool guarantees brand impact.

My strategic conclusion from this is that scaling needs a robust foundation. Which statements may vary, which must remain identical? Which information is binding? Who updates it? Who decides in cases of conflicting goals between speed, clarity and legal diligence? Without such answers, automation becomes an amplifier of existing problems.

I would therefore first strengthen the shared knowledge base and editorial guardrails. This includes verified descriptions of services, concrete language principles, traceable approvals and quality controls for sensitive content. A brand voice should not consist only of adjectives, but should use examples to show how the brand explains, disagrees and states limits. That way, technology can support expression without taking over responsibility for it.

For me, progress does not lie in saying more everywhere. It lies in being able to say the right thing in more situations: consistently, relevantly and responsibly.

If we want to assess brands in an AI context, we need to distinguish between terms. Brand awareness, put simply, describes whether people know or remember a brand. AI brand visibility, as I use the term, describes whether and where a brand appears in the answers of selected AI systems. These are different levels of observation. A frequently mentioned brand is not automatically known, well-liked, or decisive for purchase.

Even more important is a third layer: AI brand interpretation. Is the brand correctly categorized in terms of content? Does the attributed capability match the actual offer? Are relevant differences recognizable? A mention can sound positive and still rest on a false assumption. It can also simply be an incidental example. Merely counting mentions says little about this.

I therefore want to know in which need situations a brand becomes visible and what an answer communicates about it. Does it only show up for explicit name queries, or also for open category questions? Is it matched to a suitable target group? Which sources are cited? What information is missing? Only this qualitative assessment turns a list of answers into a strategically useful observation.

At the same time, I would not hastily turn AI visibility into a supposedly objective overall score. A result depends on the questions studied, the language, the timing, the product, and possibly personalization. Monitoring can describe a defined slice. It does not automatically measure the entire market, people's attitudes, or the contribution to demand.

Being visible is a precondition in certain touchpoints. Being understood correctly is another task. And meaning for people remains, once again, more than both of those combined.

For me, a single question to a chatbot is not a brand audit. It can be an interesting signal, but not a robust diagnosis. A meaningful AI brand monitoring process begins with a defined research scope: relevant needs, typical category questions, comparison situations, critical follow-up questions and different languages or markets. The selection must be justified, rather than containing only the questions in which the brand happens to look especially good.

For repeated observations, I would document which system and which identifiable version were used, when the query took place, whether search was activated, and what context applied. Repetitions help make fluctuations visible. I would evaluate not only mentions, but also accuracy, categorization, source quality and relevant omissions. Where comparisons are made over time, changes to the research setup must remain visible.

Research on Generative Engine Optimization, or GEO, offers an interesting point of reference here. The paper "GEO: Generative Engine Optimization", published in 2024, examines in an experimental setting how the visibility of sources in generative answers can be influenced by changes to content. Among other things, it looks at source references and the way information is presented. This makes the question scientifically examinable, but does not yet make every consulting approach derived from it reliable.

I would not sell results from a benchmark as a blanket guarantee for current, constantly changing assistants. Impact can vary depending on the topic area, the system and the visibility measure used. Better presence in an experiment is not proof of more revenue, brand trust, or lasting recommendation. Google's documentation, too, remains system-specific: it explains requirements for Google Search and cannot stand in for every AI offering.

My goal would therefore be a learning process, not a ranking promise. If a brand is misrepresented, we first examine the sources and our own information. If relevant content is missing, we clarify whether it is accessible and substantial enough. Monitoring is worthwhile when it leads to well-founded improvements, not when it merely turns uncertainty into a seemingly precise number.

With all this attention on systems, I do not want to lose the human measure. People do not approach brands only with an information question. They bring experiences, longings, skepticism and cultural references. A brand can be correctly described and still trigger nothing. Relevance does not arise from data organization alone, but also from a stance that is comprehensible, and from experiences that mean something.

The Adobe Creative Trends Report 2026 is, for me, a cultural impulse in this context. It describes creative directions and the role of emotional connection, expression and new design possibilities. Such trends can raise awareness and open up ideas. However, they are not a representative measurement of what all people expect from every brand, and not a guide for reliably producing authenticity.

My perspective, therefore, is not to adopt the next trend as quickly as possible. Instead, I ask which forms of expression fit a brand's identity and which merely look contemporary. When many companies use similar tools, the quality of decisions gains importance: what do we deliberately leave out? Where do we stay concrete? What real experience backs up what we communicate?

Authenticity, for me, does not mean fundamentally forgoing AI. It means that expression, origin and behavior do not diverge. A synthetic image can be a legitimate design choice, as long as it does not pretend to be a real person, experience or achievement. A personal text, by contrast, requires responsibility for its claims, regardless of which tools were involved in creating it.

Human distinctiveness is not a decorative counterweight to technology. It is what makes a brand memorable and culturally resonant at all.

The question goes even further when AI systems do not just answer, but carry out tasks. Depending on their technical design and the permissions granted, agentic applications can, for example, compare information, prepare a process, or interact with other services. Not every product marketed as an "agent" does all of this. I would therefore always first describe the actual capabilities and limits of an application.

This creates an added responsibility for brands. When a system acts on a company's behalf, its behavior becomes part of the brand experience. An incorrect price, an impermissible commitment, or an untransparent sharing of data is then not just a technical problem. People experience it as the company's behavior. A friendly tone cannot compensate for a lack of reliability.

Agent-capable brand systems therefore need current, clearly assigned information and clearly defined responsibilities. Product data, availability, contract terms and performance limits must be traceable. Equally important is the question of which decisions a system may make on its own, when explicit consent is required, and when a human must take over. I see this as a shared task for brand management, product, technology, legal and customer service.

For me, governance here does not mean preventing innovation. It creates a framework in which innovation becomes responsible: appropriate access rights, data protection, documented changes, review of critical answers and traceable paths to correction. Sensitive data does not belong in general models without controls. And a brand should not create the impression of human advice when an automated system is actually responding.

Which agentic applications will prevail in which industries remains open. I would not derive an inevitable future for every brand from this. But wherever they are deployed, a communication guideline alone is no longer sufficient. Brand identity must be translated into rules, data quality and responsible behavior.

I do not believe we should develop brands for machines instead of for people. I believe we need to understand how technology mediates between brands and people. That, for me, is precisely the new strategic task: preserving meaning while improving the conditions under which that meaning does not get lost in additional contexts.

Human Beautiful, for me, does not simply mean beautifully designed. It means understandable, distinctive, emotionally resonant and carried by genuine behavior. Machine Readable does not mean squeezing the brand into a rigid data format. It means making knowledge about it clear, accessible, current and verifiable. Both perspectives need a shared foundation: an identity that does not end at the surface.

I would therefore start with the fundamental questions. What does the brand truly stand for? Which needs does it address? What distinguishes its offer in a comprehensible way? Which claims can we substantiate? Only after that would I examine how this clarity can be translated into information, content and interactions. The order matters. Technology can reinforce a good strategy, but it cannot do the work of building its substance for us.

My own perspective also remains an invitation for further development, not a claim to the final answer. Systems change, research continues to evolve, and habits do not form everywhere at the same pace. That is why I want to observe, test and learn, without declaring every new feature a revolution and without underestimating the change underway.

For me, brands in the age of AI do not need to become less human. They need to become more deliberate about what they mean, how they are understood, and what they take responsibility for.

Human Beautiful. Machine Readable.