Interpretability is the largest pillar because it measures whether a brand has made itself understandable at all. Before a system can recommend a brand, it must be able to say, in a sentence or a paragraph, what the brand is, whom it serves, and why someone might choose it. Many brands assume this work is done. Most have not yet done it, because until recently there was little reason to.

The pillar examines every surface on which a brand explains itself: its product pages, its answers to common questions, its editorial and structured content, its collection and category pages, the way it describes itself across the site, and the consistency of all of these with one another. The question throughout is whether the brand reads coherently when a system encounters it, whether its descriptions match its substance, and whether what the brand says about itself is what a careful reader would conclude after examining its work.

This is harder than it sounds, and it is rarely a matter of effort. A brand can have beautifully designed pages, abundant content, and a clear visual identity and still not be understood. The most common gap is between atmosphere and explanation: a page built to be felt rather than read communicates a feeling without committing to what the product does, whom it serves, or what distinguishes it from the things beside it. This is not a fault. For most of the web’s history, a page’s job was to be seen, and a page that evokes the right feeling did that job well. But a system constructing a recommendation does not feel a page; it reads it. It can register the brand’s mood and still be unable to recommend the brand with confidence, because it cannot say the conditions under which the recommendation would be right.

The remedy is not to say more. It is to ensure that what the brand means is legible beneath what it evokes. The most interpretable brands in luxury and wellness are often the most atmospherically assured, because they have understood that precision about substance is what gives atmosphere somewhere to stand. Their categories are clear, their audience is defined without hedging, and the occasions on which they are the right answer can be known, none of which costs them their restraint. For brands whose meaning is culturally rooted, this is also how native codes become legible to a system that may not yet share them: not by explaining the mystery away, but by ensuring the substance beneath it can be read.

Trust Signals measures the evidence that surrounds a brand: what others say about it, what it can demonstrate about its claims, and what the wider environment communicates that the brand cannot say about itself. It is the pillar most closely tied to whether a system will recommend a brand with confidence. A system does not believe a brand because the brand asserts it should be believed. It believes the brand when the surrounding evidence converges on the same account.

The pillar examines the reviews a brand hosts on its own surfaces, the reviews and discussion that accumulate on third-party platforms, the language of the social and user-generated content around it, the press and editorial and affiliate context in which it appears, and the authority signals attached to the people, expertise, or institutions behind it. Each of these adds to, or subtracts from, the certainty with which a system can recommend the brand. A brand with strong first-party reviews and no outside corroboration reads as self-described. A brand with abundant press and no lived discussion reads as marketed rather than chosen. The signals have to agree for the recommendation to feel earned.

One dimension within this pillar deserves to be named directly, because it is among the things Lunera weighs that most frameworks overlook: the evidence of experience itself. A system does not only read what is written about a brand. It reads images, video, and user-generated content, and the systems that mediate discovery are growing steadily more capable of drawing meaning from them. This capability is not yet complete, but its direction is not in question, and it is moving toward a point where the coherence between what a brand claims and what its lived evidence shows will be unavoidable to any system constructing a recommendation. Lunera applies this lens now, before that moment arrives, because the work of aligning claim and evidence takes time, and the brands that begin while the cost of incoherence is still low will be ready when it is no longer optional. For a hotel, whether the light in a guest’s photographs accords with the atmosphere the brand describes; for a wellness brand, whether the people and settings in circulation reinforce the ritual it claims; for a luxury house, whether the contexts in which its work appears confirm its register. These are the kinds of coherence a maturing system will increasingly read, and the kind a brand is wise to establish before it must.

A second direction is also worth naming, because it is just beginning to appear in the surfaces that mediate discovery. The systems are starting to incorporate signals of declared trust — the publications a person has chosen to follow, the sources they have asked to see more of, the relationships of repeat readership and subscription that accumulate over time. These are early signals, currently scoped to particular surfaces and uses, and it would be premature to treat them as decisive. But the trajectory is consistent with everything else in this pillar: that trust assembled across time and across independent parties is what allows a system to recommend with confidence. As declared-preference signals expand, the brands that have earned a place in their audience’s deliberate attention — not only their incidental visits — will hold a kind of evidence that algorithms alone cannot supply and that competitors cannot easily fabricate.

This is also why trust cannot simply be purchased or produced. A brand can commission praise, but it cannot easily commission the convergence of independent evidence, and convergence is what a system reads. The brands that score well here are not the ones that have said the most about themselves. They are the ones whose own account, the customer’s voice, the press treatment, and the lived evidence of experience all describe the same thing. The confidence a system places in such a brand is not granted. It is the natural result of evidence that agrees.

AI Visibility measures whether a brand actually surfaces in AI-mediated discovery, and how accurately it is represented when it does. It is the most directly testable of the pillars, because its evidence comes from putting questions to the systems themselves and observing what they return. The other pillars assess the conditions of being understood; this one assesses the result.

The pillar examines a brand’s presence across several kinds of inquiry. Direct queries, in which a brand is named, test whether the system recognizes it and describes it correctly. Comparative queries, in which a brand is asked to be weighed against alternatives, test whether the system places it in the right company and renders its distinctions accurately. Problem-oriented queries, in which a brand is not named but a need is described, test whether the system surfaces the brand when it is the right answer and not when it is not. Each kind of inquiry reveals something different. Together they describe both whether a brand can be found by an AI system and how it is understood once it is.

This is the pillar in which the two halves of being understood most plainly meet. A brand may surface in every relevant query and still be misrepresented in the answers. It may be described accurately when named directly and yet placed beside the wrong competitors in comparisons. It may be invisible in one system and prominently recommended in another. These are different conditions with different causes, and the pillar treats them distinctly: surfacing is one question, accurate description is another, and consistency across systems is a third. A brand cannot be assumed to be visible because it appears in one place, nor accurately understood because it appears at all.

For premium brands, the comparative dimension carries particular weight. A luxury house is not chosen against everything; it is chosen against its peers, and the company it is held to keep is part of what it is. When a system, asked to compare brands, places a maison among houses of a different register, or attributes to it the language of an adjacent but distinct category, the harm is not generic. It is the precise harm of being weighed in the wrong room. Premium brands depend on being correctly situated, and AI Visibility is where situation is observed.

The pillar’s evidence is gathered through structured inquiry across the systems that now mediate discovery, with attention to what each returns, how each describes the brand, and where the answers agree and diverge. The brands that score well here are not necessarily those that appear most often. They are the ones whose appearances are accurate, whose representations across systems converge, and whose presence is consistent with the position the brand has worked to hold.

Semantic Structure measures the technical foundation that allows AI systems to reach a brand at all and to read it once they arrive. It is the substrate beneath every other pillar. A brand cannot be understood by a system that cannot access it, and cannot be accurately described by one that cannot parse what it sees. The pillar carries less weight than those above it because the other pillars contend with what is harder to fix; this one contends with what is necessary but largely solvable.

The pillar examines whether a brand’s content is reachable by the systems that crawl and read it, whether its pages render in forms those systems can interpret, whether its underlying markup describes its content accurately, whether its metadata says what it appears to say, whether its language and regional structures resolve cleanly, and whether its architecture avoids the fragmentation that fractures a single brand into many partial versions of itself. These are unglamorous concerns, often treated as the province of technical specialists who do not need to think about brand strategy. The Index treats them differently. Technical legibility is what allows the work of being understood to be done at all. A brand cannot interpretively cohere across surfaces that a system cannot consistently read.

Most of what this pillar measures was, until recently, well within the discipline of search engineering, and many brands have already done much of this work. What is changing is what the systems on the other end are looking for, and how strictly they hold a brand to what its structure says about itself. A page that renders inconsistently for different readers, or claims one thing in its markup and another in its content, no longer simply ranks slightly lower. It becomes harder to trust as a coherent account. The technical work is no longer only about ranking. It is about whether the brand’s signals agree with themselves at the level beneath the words.

For brands with international presence, this pillar carries an additional weight. A maison with European headquarters, an American showroom, regional retailers, and a separate site for each market will, without deliberate management, present a system with several plausible versions of itself, none of which fully resolves. The technical structures that distinguish primary from regional, current from archived, canonical from variant, are not minor administrative concerns. They are the architecture by which a single brand remains a single brand to the systems that read it.

The pillar’s weight is appropriate to its nature. The work it measures is foundational rather than expressive, and most of it can be remediated in months rather than years. But its weight is also not negligible. Strong work in every other pillar can be silently undermined here, and a brand that overlooks the foundation will find that its more visible investments yield less than they should.