AI Predicts Where Shoppers Look in Luxury Beauty Stores Before They Step Inside

…

…

The study’s methodological foundation rests on a growing class of AI platforms that attempt to reconstruct human visual and neural responses from large aggregated datasets rather than measuring them directly in a given store. In this case, the predictive model drew on eye-tracking data from roughly 180,000 participants and EEG recordings from 27,000 participants, amounting to more than 100 billion data points. Such datasets allow algorithms to learn statistical regularities in how human gaze and neural activity respond to visual features—contrast, color, density, faces, text, and spatial arrangement—and then apply those regularities to new stimuli. The critical distinction, which the authors make explicit, is that no shoppers were actually tested in the Zagreb cosmetics store. Every reported value is a model prediction, which places the work in the category of exploratory simulation rather than confirmatory field experimentation.

This distinction matters for interpretation. Traditional in-store eye tracking requires recruiting participants, equipping them with mobile tracking hardware, and navigating ethical and logistical constraints of filming real shoppers. Predictive platforms bypass these hurdles, offering rapid, low-cost evaluation of layouts and displays before physical changes are made. The trade-off is that predictions inherit the limitations of their training data, which may underrepresent specific populations, lighting conditions, or cultural contexts, and they cannot capture idiosyncratic motivations, prior brand relationships, or purchase intent that only real shoppers bring. The authors frame their findings accordingly: as hypotheses and design propositions that would need validation with live shoppers before being treated as settled guidance.

The choice of stimuli reflects a deliberate two-tier design. From 21 photographs of the store, 11 static images were selected for frame-by-frame analysis, allowing areas of interest to be defined on the elements the researchers considered critical in each view—brand zones, product displays, counters, and competing visual material. Separately, a 9-minute continuous in-store video was edited down to an 85-second simulation of the consumer journey, compressing the walkthrough into a dynamic sequence that could be treated as a time series. This dual approach lets the researchers examine both spatial patterns, where attention concentrates across the store at a given moment, and temporal patterns, how predicted attention and cognitive demand evolve as a shopper moves through the space.

Among the static findings, the strongest and most consistent pattern concerns visual separation. Brand areas that stood apart from competing stimuli attracted stronger and more sustained predicted attention, while visually crowded zones and areas exposed to external distractions—such as views of neighboring stores—showed reduced attentional efficiency. This aligns with long-standing principles from visual attention research: clutter increases the competition for limited attentional resources, and salient elements lose their advantage when surrounded by equally salient neighbors. In a luxury context, where brands invest heavily in distinctive display design, the finding suggests that the protective space around a display may be as valuable as the display itself.

Central product placement emerged as particularly effective, but with an important qualifier: it worked best when surrounded by fewer distracting elements. This refines the familiar retail maxim that eye level is buy level. Visibility at eye level appears to be a necessary but insufficient condition; the surrounding visual field determines whether that visibility translates into efficient processing. A centrally placed product embedded in a dense, competing visual environment may attract a glance without sustaining attention, whereas the same placement in a visually quiet zone supports the longer, more focused gaze that previous research has linked to choice probability.

The correlational results at the area-of-interest level add a layer of nuance. Across the defined AOIs, Engagement and Cognitive Demand rose together with a correlation of 0.90, a strikingly strong relationship. In practical terms, the elements of the store that drew the most predicted engagement were also the ones demanding the most cognitive effort. This could reflect the nature of luxury beauty retail itself, where products, ingredients, and brand stories invite close inspection and comparison, or it could indicate that visually rich displays are inherently both attractive and demanding. Either way, the coupling challenges the assumption that engagement and ease of processing always go hand in hand; in this environment they appear to be two sides of the same stimulus properties.

The video simulation told a different story. As the 85-second walkthrough unfolded, Cognitive Demand remained uniformly high throughout the journey, while Focus varied with the field of view—and, notably, the two measures were unrelated in the dynamic condition. This dissociation between static and dynamic results is one of the study’s most interesting methodological observations. A single photograph freezes the visual field and lets the model evaluate a fixed composition; a moving walkthrough constantly changes what is visible, introducing new displays, angles, and competing storefronts. The uniform cognitive demand across the video suggests that navigating a visually dense luxury store keeps mental effort elevated regardless of where the shopper looks, while the fluctuating focus indicates that attention is captured and released in bursts tied to specific views.

The comparison of views with and without competing stores in the field of view deserves particular attention. Windows and sightlines that exposed the store’s interior to neighboring retail environments reduced attentional efficiency for the focal brand zones. For store designers, this implies that sightline management—through window treatments, interior partitions, or display placement that blocks external views—may be a controllable lever for protecting in-store attention, especially in shopping districts where adjacent storefronts compete aggressively for visual attention.

The analytical framework itself constitutes a contribution beyond the specific findings. The authors describe a repeatable procedure: select representative views, define areas of interest on the critical elements of each view, correlate the predicted metrics across areas and views, compare views with and without competing visual material, and treat the walkthrough as a time series. Because the pipeline relies on predicted rather than measured data, it can be applied repeatedly and cheaply, including to hypothetical layouts that exist only as renderings. This opens the possibility of testing design propositions before a store is built or refitted—evaluating, for instance, whether a proposed display configuration would generate the spatial clarity associated with efficient visual processing, before committing to construction costs.

For communication research, the study extends visual communication theory into immersive retail spaces by following predicted attention through an entire store and its surroundings rather than isolating single stimuli. Retail environments differ fundamentally from advertisements or web pages: they are navigable, three-dimensional, and filled with overlapping brand messages, and the attention patterns they produce cannot be inferred from flat media. By treating the store as a communication medium in its own right, the study positions layout, lighting, and spatial hierarchy as communicative acts that shape what shoppers see, how hard they work to see it, and by extension how brand messages are processed.

The practical implications for luxury beauty retail follow directly from the patterns observed. Displays should be given breathing room, with visual separation from competing stimuli treated as a design resource. Eye-level placement should be paired with reduced surrounding clutter to convert visibility into sustained attention. Sightlines to the outside world should be managed where external exposure degrades attentional efficiency. And because cognitive demand appears uniformly high across the walkthrough, retailers may need to consider whether that sustained effort is desirable—supporting deep engagement with premium products—or whether moments of visual rest should be deliberately designed into the journey to prevent overload.

Several limitations temper these conclusions. The single-store setting in Zagreb limits generalizability to other luxury retail formats, cultures, and market contexts. The sample of 11 static images and one edited video, however systematically chosen, cannot capture the full variability of the retail environment or of shopper paths through it. The model predictions, however well calibrated against their training datasets, remain statistical estimates rather than observations of actual shopper behavior, and the study’s exploratory design means its research questions were framed to generate hypotheses rather than test them. The strong Engagement–Cognitive Demand correlation, in particular, warrants follow-up with measured data to determine whether the coupling holds when real shoppers, with real goals and time pressures, move through comparable spaces.

Future research directions follow naturally from these limitations. Validation studies combining predictive outputs with live mobile eye tracking and EEG in the same stores would establish how well the predictions correspond to measured attention and neural effort. Cross-store and cross-category comparisons would reveal whether the observed patterns—spatial clarity benefits, the engagement–demand coupling, the dissociation in dynamic conditions—are specific to luxury beauty retail or generalize across retail formats. Longitudinal designs could test whether layouts optimized through prediction actually produce measurable differences in attention, dwell time, and sales. And extending the framework to other sensory channels, since luxury retail is deliberately multisensory, would connect predicted visual attention to the broader experiential fabric of scent, sound, and touch that defines premium brand environments.

Within the broader trajectory of neuromarketing, the study illustrates a shift from measuring small samples of consumers in laboratory conditions to predicting population-level responses from aggregated big data. This shift brings both opportunity and responsibility: opportunity, because design iteration becomes faster and cheaper; responsibility, because the provenance and representativeness of training data, and the transparency of prediction limits, become central to the credibility of any conclusions drawn. The authors’ explicit framing of their results as predictions, and their careful separation of predicted metrics from measured behavior, offer a model for how such platforms can be reported in academic and applied contexts alike.

Taken together, the findings sketch a coherent picture of the luxury beauty store as a visual communication system in which attention is a scarce resource shaped by spatial organization. Brand zones that are visually protected, centrally placed, and surrounded by clarity capture and hold predicted attention; crowded and externally exposed zones squander it. Cognitive demand runs high throughout the shopping journey and tracks engagement in static views, suggesting that luxury retail communication is inherently effortful. For retailers, the message is that the architecture of attention—where displays sit, what surrounds them, and what the shopper can see beyond them—is not decoration but a measurable, optimizable component of brand communication, one that can now be evaluated computationally before a single shelf is installed.

Subject of Research: …

Article Title: Predicted visual attention and cognitive demand in luxury beauty retail: an exploratory AI-based neuromarketing study

Article References: Šola, H. M., Gajdoš Kljusurić, J., & Šestanović, A. (2026). Predicted visual attention and cognitive demand in luxury beauty retail: an exploratory AI-based neuromarketing study. Frontiers in Communication, 11, Article 1933317. https://doi.org/10.3389/fcomm.2026.1933317

Image Credits: AI Generated

DOI: 10.3389/fcomm.2026.1933317

Keywords: neuromarketing, eye tracking, EEG, artificial intelligence, luxury retail, cognitive demand, visual attention, store layout, consumer behavior, retail design, Predicted, visual