As large language models (LLMs) increasingly influence consumers’ product discovery and purchase decisions, unpacking how brands are “seen” and recommended – what some call bot psychology – by AI gatekeepers has become critical. Many executives are asking: "How do we get our brand recommended by AI?" A more important question may be: "Does AI understand what makes our brand valuable?" 

Evidence suggests that AI systems value utilitarian and explicit cues in interpreting meaning and generating answers to consumers’ prompts. Google's own recently published guidance suggests that in order to be seen by LLMs, brands should create clear and compelling content within a legible structure. 

But this goes against the long-standing playbook of many aspiration and luxury brands, from sports to fashion and even high-end hospitality. Such brands tend to tap symbolic and implicit associations rather than playing up explicit, functional attributes. Cultural schema (“artistic association is valuable”), shared assumptions (“higher means greater status”), embodied metaphors (“thin bottles signal elegance”) and subtle cues (e.g. slow pacing of ads) are used to convey scarcity and heritage to humans, not what can be read at face value. 

Can AI systems correctly interpret such intuitive and often subconscious cues, read between the lines, and integrate the unsaid? As we wrote in a recent article in Harvard Business Review, the likely answer, based on a series of studies, is “no”.

Testing AI brand desirability

Our experiments systematically tested AI's response to four well-documented luxury cues known to boost products’ desirability: higher physical positioning, association with art, spacious display and slender design and packaging. We used original or near-identical stimuli from prior studies to allow direct comparison. Three models were tested: ChatGPT 5.1, Claude Sonnet 4.5 and Gemini 3 Pro. We sampled each model 150 times for each stimulus, which included chocolates, jewellery, tableware, beauty products, watches and more.

The models handled explicit cues such as a stated brand name and the word "luxury" well enough. Everything else was another matter. Unlike humans, the LLMs didn’t prefer products placed at a higher physical position or presented in a more spacious setting (see charts below). Art associations had less impact on LLMs than on humans, and AI models seemed to prefer photographs – especially those of celebrities popular on YouTube or Reddit – over paintings.

In short, shape- and proportion-based cues barely registered. Minimalist, spacious visual environments produced negative responses. More white space was correlated with lower perceived value. 

Our experiment indicates that visibility and recognition play a central role in shaping luxury perceptions among LLMs compared to humans, but algorithms misinterpret luxury attributes and mischaracterise what makes them desirable. The implicit, hedonic cues that luxury brands have relied on for decades to build desire can lead to weaker, not stronger, visibility in AI-generated responses. 

How brand context shifts AI judgment

In our second experiment, the three LLMs sized up six car brands – Alfa Romeo, BMW, Ferrari, Mercedes, Porsche and Tesla – against either a plain background or a luxurious one featuring a gilded-framed Van Gogh painting. In all, we generated 5,400 evaluations of the LLMs’ willingness to pay for each brand.

The results revealed that Mercedes benefited from the luxury context across all three models whereas Porsche was penalised. Ferrari, meanwhile, divided opinion: ChatGPT assigned it lower value in the luxury context, Gemini was unmoved and Claude valued it more highly. Apparently, for brands whose identity is built around performance rather than heritage, showing up in a luxury context appears to create resistance rather than endorsement.

Two further takeaways stand out. First, consumers’ intuitive understanding that a Ferrari occupies a different category from a BMW doesn’t register with AI unless it’s made explicit, such as through rankings, tier labels (e.g. “premium” or “luxury”), comparison tables and awards. Otherwise, models may treat brands far apart on the luxury hierarchy as equally prestigious, which impacts which brands get surfaced and recommended.

Second, the models apply different interpretive lenses (see chart below) to the same material. A content strategy calibrated for one model can backfire on another. Brands need to test the same asset across multiple LLMs, identify where models agree and where they diverge, and adjust accordingly.

A new luxury marketing playbook

Luxury marketing was built on human psychology, but the future will increasingly require that brands design for machine psychology as well. Luxury leaders must close the meaning gap. Rethinking strategy through the famous 4Ps of marketing – product, price, place and promotion – provides a useful organising frame.

Product: AI systems favour explicit descriptors over implicit cues. This means craftsmanship, provenance, materials and design intent must be articulated. Brand executives should audit their asset inventory of campaign imagery, product descriptions, taglines and visual language, and assess each for AI readiness. Ask yourself: How much meaning would survive if the implicit signals were removed? 

Audit your brand through the eyes of AI. Many organisations have mapped customer journeys to the last detail but have little understanding of how AI currently describes, categorises and recommends their brand. How does AI position your brand vs. your competitors? Does it classify you as luxury, high-end, innovative, value-oriented or something else entirely? Where is it getting that information from?

An AI context strategy brief, developed alongside traditional brand guidelines, should specify how products should be described, in what use contexts, and relative to which competitor set. A jewellery brand, for instance, should leverage engagements, anniversaries and significant personal milestones to give models the context they need to infer value.

Price: Willingness-to-pay experiments give brand managers a practical monitoring tool. If one model labels a product "overpriced" while another calls it "premium", it’s a signal that the cues surrounding the product aren’t sufficient for the model to infer intended positioning. Identifying where AI systematically undervalues a brand or inflates a competitor enables executives to make corrections before wrong or inadequate information shapes consumer decisions.

Promotion: Luxury brands risk losing control of meaning when functional product attributes are left ambiguous. In a separate analysis of ski brand Atomic, we found that AI interpreted the rigidity of the brand's skis – a core performance attribute – as a drawback. 

The solution is to probe the models to understand what descriptors they currently attach to your brand. Define the intended reading, identify the gaps, and use precise, high-status language consistently across owned and earned channels to close them. Third-party content, including reviews, mass-market comparisons and off-brand associations, continues to be indexed long after a brand has moved on strategically. Managers should conduct a comprehensive audit of existing content and ensure that AI retrieves the most accurate information about the brand when consumers ask for guidance. 

Placement: Brands’ owned websites are only a fraction of the equation. Our research (presented in the chart below) shows that roughly 80% of an LLM’s citations come from third-party e-commerce platforms, news media, blogs, Reddit threads and YouTube videos. This broader ecosystem is the new frontline for positioning your brand. In an AI-mediated market, visibility is about more than just keywords – it's about understanding machine psychology to ensure that meaning is never lost in translation. This understanding entails treating the content ecosystem holistically, and tailoring content strategies to the algorithms of different platforms.

A new mandate for brand leaders

Luxury brands have spent decades building what might be called a visual vocabulary that tells a consumer, without spelling it out, that their product is different from the rest. AI systems aren’t immersed in that vocabulary. Luxury brands’ pressing task in the age of LLMs is to find the layer of explicit language that preserves meaning without dissolving the mystique, making the brand readable without rendering it ordinary. 

This article is adapted from “LLMs Misunderstand Luxury Brands. Here’s How to Optimize Your Content Strategy for AI” published in Harvard Business Review.