{"id":79860,"date":"2026-11-06T17:33:19","date_gmt":"2026-11-06T22:33:19","guid":{"rendered":"https:\/\/digital.hec.ca\/?p=79860"},"modified":"2026-11-06T17:33:21","modified_gmt":"2026-11-06T22:33:21","slug":"how-ai-shopping-assistants-choose-brands","status":"publish","type":"post","link":"https:\/\/digital.hec.ca\/en\/blog\/how-ai-shopping-assistants-choose-brands\/","title":{"rendered":"When AI Makes the Shortlist: Why Brand Positioning Is Important Now More Than Ever"},"content":{"rendered":"\n

What more than 50 shopping shortlists across ChatGPT, Gemini and Perplexity reveal about brand visibility<\/em><\/p>\n\n\n\n

Branding is one of the main tasks of digital marketing. In order to promote a product or service, a company invests in advertising, search rankings, social media and content marketing to ensure that potential customers know about it.<\/p>\n\n\n\n

The logic is quite simple: if customers do not know about the existence of a brand, they cannot select it.<\/p>\n\n\n\n

However, what happens when consumers do not try to find a product themselves but ask an AI assistant to provide some options?<\/p>\n\n\n\n

I think it is a relevant issue for marketing professionals. A company can have a popular brand, high-quality products and a strong web presence. Still, that does not mean that an AI assistant will include the company among its recommendations. Consumer needs can influence brand recommendations.<\/p>\n\n\n\n

As a Management and Artificial Intelligence student at LUISS, currently on exchange at HEC Montr\u00e9al, I wanted to see the connection between AI shopping assistants and traditional marketing concepts. I tested shopping recommendations from three AI assistants for different product categories to see whether specifying consumers’ needs could affect the recommendations.<\/p>\n\n\n\n

This work allowed me to rethink the concept of brand positioning and brand visibility in general.<\/p>\n\n\n\n

1. AI is changing the way that consumers come into contact with brands.<\/h2>\n\n\n\n


The classic process of an online journey of the consumer involves many steps. The consumer searches for the product, compares it, studies the reviews, goes to physical stores and, in the end, chooses the product. There are various chances to attract the attention of the consumer during the whole process.<\/p>\n\n\n\n

    <\/ol>\n\n\n\n

    AI shopping assistants provide another option for the consumer. Instead of visiting five websites to compare different models of headphones, a student can ask ChatGPT to recommend the best models under CAD 300 for studying in a noisy environment.<\/p>\n\n\n\n

    And the assistant compiles a list according to the request.<\/p>\n\n\n\n

    That is important because a consumer may never encounter many brands that would otherwise appear in traditional search results.<\/p>\n\n\n\n

    There are signs that AI-assisted shopping is becoming commercially relevant. According to Adobe (2026)<\/a>, traffic from AI sources to websites of American retailers increased by 393% year over year in Q1 2026. In March, the conversion rate of visitors who came through AI sources was 42% higher compared with other visitors.<\/p>\n\n\n\n

    It does not confirm that the recommendations of AI assistants increase conversion rates. Even so, it suggests that AI-assisted shopping is growing in commercial importance.<\/p>\n\n\n\n

    According to a survey conducted by Accenture in 2026<\/a>, consumers distinguish between different levels of trust in AI shopping assistants. While 32% of respondents would let an AI agent choose products and leave the payment to the consumer, only 9% would allow autonomous purchasing within predefined limits.<\/p>\n\n\n\n

    For now, I find it more relevant how AI-assisted shopping can change the discovery phase than how it can make autonomous purchasing decisions.<\/p>\n\n\n\n

    2. Brand positioning is still important, just in another setting<\/h2>\n\n\n\n

    One of the first marketing concepts that a student learns is the concept of brand positioning. A brand creates an image of itself in relation to competitors.<\/p>\n\n\n\n

    Nike creates its image on the basis of athletic performance and innovative technologies. Apple is associated with design, minimalism and its product ecosystem. That is what makes customers recognize these brands.<\/p>\n\n\n\n

    In AI-assisted shopping, however, brand positioning can perform another task. A consumer describes their needs, and the AI assistant has to identify brands that can potentially satisfy those needs.<\/p>\n\n\n\n

    That is when Generative Engine Optimization (GEO) comes into play.<\/p>\n\n\n\n

    GEO refers to improving the visibility and usefulness of content in the answers provided by AI assistants. This concept was introduced in academic research by Aggarwal and colleagues (2024)<\/a>.<\/p>\n\n\n\n

    As discussed in From SEO to GEO<\/a> on this blog, brands also need to think about visibility beyond traditional search results.<\/p>\n\n\n\n

    Unlike traditional SEO, where marketers analyze website rankings and clicks in search engines, GEO raises the question of whether a brand or information about it is present in an AI-generated answer.<\/p>\n\n\n\n

    Key terms: SEO, GEO and AI visibility<\/h3>\n\n\n\n

    Several terms can be helpful in understanding the concept:<\/p>\n\n\n\n

    Term<\/th>Meaning<\/th><\/tr><\/thead>
    Search Engine Optimization (SEO)<\/td>Improving the visibility of a brand’s web resources in search engine rankings.<\/td><\/tr>
    Generative Engine Optimization (GEO)<\/td>Improving the availability, relevance and credibility of content that can be used in AI assistants’ answers.<\/td><\/tr>
    AI shopping assistant<\/td>An AI system that helps consumers discover, compare and evaluate products.<\/td><\/tr>
    Brand Recommendation Probability (BRP@5)<\/td>The percentage of responses in which a brand is included in the shortlist of five recommended options.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    The key question is whether the advantage of well-recognized brands remains relevant when consumers describe their specific needs.<\/p>\n\n\n\n

    For example, will the most recognized brands of running shoes remain in the lead if a consumer looks for shoes suitable for wide feet?<\/p>\n\n\n\n

    Without trying to predict the answer, I decided to check it myself and test the recommendations of three assistants.<\/p>\n\n\n\n

    3. Comparison of brand recommendations in three AI assistants<\/h2>\n\n\n\n

    For this experiment, I selected three product categories that would be interesting for university students: laptops, wireless headphones and running shoes.<\/p>\n\n\n\n

    In each case, I compared a general shopping request with a request that included a specific consumer need.<\/p>\n\n\n\n

    Product category<\/th>Budget<\/th>Additional consumer need<\/th><\/tr><\/thead>
    Laptops for university students<\/td>Under CAD 1,500<\/td>Battery life lasting a full day of classes<\/td><\/tr>
    Wireless headphones<\/td>Under CAD 300<\/td>Noise cancellation for studying in a noisy environment<\/td><\/tr>
    Running shoes<\/td>Under CAD 200<\/td>A wide fit for wide feet<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    The general request asked for a ranked list of five recommendations, including the brand and model. The second version included the same request and an additional sentence that described the consumer need.<\/p>\n\n\n\n

    I repeated both requests three times with ChatGPT, Gemini and Perplexity, resulting in 54 responses between October 15 and 25, 2026. For the experiment, I used temporary or private chat sessions on free accounts to reduce the personalization effect and recorded the brands included in the recommendations, their positions and references to sources, if any.<\/p>\n\n\n\n

    That was not a representative study of AI shopping behavior. I wanted to see whether famous brands would dominate the general recommendations and how more specific needs would change the shortlists.<\/p>\n\n\n\n

    What happens when the consumer becomes more specific?<\/h3>\n\n\n\n

    The first thing I noticed was that some well-known brands were included in all recommendations.<\/p>\n\n\n\n

    Apple and ASUS appeared in all laptop recommendations. Sony and Anker Soundcore appeared in all headphone recommendations.<\/p>\n\n\n\n

    Meanwhile, some other brands started appearing more frequently when the question included a specific need.<\/p>\n\n\n\n

    HP appeared in only one of nine general recommendation lists but in six of nine need-specific lists when battery life became the priority. Bose appeared in four lists instead of one when I mentioned noise cancellation.<\/p>\n\n\n\n

    Altra, a running-shoe brand known for its wide, foot-shaped toe box<\/a>, did not appear in any general running- shoe recommendations but appeared in four of nine need-specific lists when I mentioned wide feet.<\/p>\n\n\n\n

    These observations suggest that the inclusion of a brand in AI assistants’ recommendations may depend not only on brand recognition but also on how closely the product matches the needs of the consumer.<\/p>\n\n\n\n

    I call these patterns default visibility and conditional visibility. Default visibility means the appearance of a brand in general recommendations without mentioning any specific need. Conditional visibility means increased visibility of a brand when a specific need is mentioned.<\/p>\n\n\n\n

    These are labels that I used to interpret the results of the experiment, not categories used by the assistants.<\/p>\n\n\n\n

    4. The Nike case: why having a particular feature is not enough<\/h2>\n\n\n\n

    The category of running shoes showed the biggest difference.<\/p>\n\n\n\n

    With the general question, Nike appeared in eight of nine lists and Adidas in six. In the case of the need- specific question, neither of these brands was included in any of the nine lists.<\/p>\n\n\n\n

    At the same time, New Balance went from four general lists to nine need-specific lists, while Brooks appeared in all need-specific shortlists.<\/p>\n\n\n\n

    New Balance also ranked first in eight of nine need-specific lists.<\/p>\n\n\n\n

    The difference can be observed across all three AI assistants:<\/p>\n\n\n\n

    AI assistant<\/th>Nike: general \u2192 wide fit<\/th>New Balance: general \u2192 wide fit<\/th><\/tr><\/thead>
    ChatGPT<\/td>3 \u2192 0<\/td>2 \u2192 3<\/td><\/tr>
    Gemini<\/td>3 \u2192 0<\/td>1 \u2192 3<\/td><\/tr>
    Perplexity<\/td>2 \u2192 0<\/td>1 \u2192 3<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
    \"Dot<\/figure>\n\n\n\n

    Figure 1. Frequency of brand appearances in the top five, generic versus need-specific prompts (9 responses per condition). Source: author\u2019s dataset, October 2026.<\/em><\/p>\n\n\n\n

    Having a feature vs. being associated with it<\/h3>\n\n\n\n

    What stands out in this case is that Nike has offered running shoes in extra-wide sizes<\/a>, including versions of its Pegasus line.<\/p>\n\n\n\n

    Therefore, the issue is not necessarily the lack of a product with a particular feature. The question is why other brands appeared more frequently when the consumer mentioned this need.<\/p>\n\n\n\n

    Of course, I cannot say anything definite about the reasons for the observed pattern. AI assistants do not disclose all factors they use to make recommendations.<\/p>\n\n\n\n

    One possible explanation, though, is that certain brands are more strongly associated with this need in the information available to the assistants. Even several of the items in the list had “Wide” in their titles like New Balance Fresh Foam X 880v15 Wide and Brooks Ghost 17 Wide. <\/p>\n\n\n\n

    In terms of marketing, it makes a significant difference whether you are marketing a product with a specific characteristic or whether you are marketing your brand based on the characteristic. It does not matter how good your product is if you are unable to properly market its characteristics.<\/p>\n\n\n\n

    That is not entirely new to marketing. Companies have always been trying to associate themselves with particular consumer needs. With AI, these associations may also matter when the system creates a shortlist.<\/p>\n\n\n\n

    5. Where do AI assistants get the information?<\/h2>\n\n\n\n

    Another question is where AI assistants get their information.<\/p>\n\n\n\n

    Product recommendations can be based on several sources of information: brand websites, retailers, independent reviews and others. The process of gathering information differs between systems and is not entirely transparent.<\/p>\n\n\n\n

    I examined the sources cited by Perplexity, as it displayed more references in its answers than ChatGPT and Gemini in my experiment.<\/p>\n\n\n\n

    In my experiment, the number of recorded references increased from 299 in the case of general questions to 486 in the case of need-specific ones. The proportion of references from brand-owned websites also increased from 17% to 29%.<\/p>\n\n\n\n

    Retailers remained the biggest source category: 58% in the case of general prompts and 50% in the case of need-specific ones.<\/p>\n\n\n\n

    \"Stacked<\/figure>\n\n\n\n

    Figure 2. Perplexity\u2019s recorded source-reference types, generic versus need-specific prompts. Source: author\u2019s dataset, October 2026; classification based on displayed references.<\/em><\/p>\n\n\n\n

    These figures should be treated cautiously. These are sources cited in the answers, but they do not represent the whole picture of the information gathered by the AI assistant. The source log does not contain every original URL needed for independent verification.<\/p>\n\n\n\n

    Why clear product information matters<\/h3>\n\n\n\n

    Still, the pattern raises a practical question for digital marketing managers.<\/p>\n\n\n\n

    When consumers ask about a specific feature of a product, like battery life or shoe sizes, does the quality of the information about the product become especially important?<\/p>\n\n\n\n

    I think it is a reasonable possibility, although my experiment does not prove a causal relationship.<\/p>\n\n\n\n

    According to research by Adobe in 2026<\/a>, the average machine readability of websites was lower for product pages than for homepages: 66% and 75%, respectively. It is important because product specifications are exactly what consumers need when they compare different alternatives.<\/p>\n\n\n\n

    Therefore, one element of a marketing strategy in the context of AI-assisted shopping can be the clear presentation of the product’s key features.<\/p>\n\n\n\n

    6. Implications for digital marketing strategies<\/h2>\n\n\n\n

    The main thing I can conclude from the results of my experiment is that companies should pay more attention to the connection between their positioning and the questions that consumers may ask.<\/p>\n\n\n\n

    Start with specific consumer needs<\/h3>\n\n\n\n

    First of all, a company should understand what problems it solves exceptionally well.<\/p>\n\n\n\n

    Instead of focusing on general notions such as “the best running shoes,” a brand can focus on specific features such as comfort, wide feet or long-distance running.<\/p>\n\n\n\n

    That links GEO with traditional market segmentation and positioning. Different consumers have different needs, and those needs can influence the recommendations they receive.<\/p>\n\n\n\n

    Make the product information clear<\/h3>\n\n\n\n

    Product pages should clearly communicate the advantages of the products.<\/p>\n\n\n\n

    In the case of laptops, it is battery performance. For running shoes, available sizes and fit. For headphones, noise cancellation and other technical characteristics.<\/p>\n\n\n\n

    The same characteristics should be present on retailers’ websites.<\/p>\n\n\n\n

    According to Google Search Central documentation<\/a>, no special GEO-specific tags are required for websites to appear in Google’s AI search features. Clear and accessible content remains important for SEO.<\/p>\n\n\n\n

    Monitor brand recommendations for different questions<\/h3>\n\n\n\n

    Digital marketing specialists track search rankings, traffic and conversions. The recommendations of AI assistants may require another way of measuring results.<\/p>\n\n\n\n

    It is not just about whether ChatGPT recommends the product or not. A company can compare recommendations for several different questions related to consumers’ needs.<\/p>\n\n\n\n

    Repeated testing is also important.<\/p>\n\n\n\n

    In comparisons of repeated answers within the same test conditions, none produced exactly the same five brands in the same order. Some brands appeared more frequently, but their positions and other recommendations changed.<\/p>\n\n\n\n

    It means that one screenshot is not a good indicator of a brand’s AI visibility.<\/p>\n\n\n\n

    A more meaningful approach would be to test several assistants, repeat the prompts and analyze how the results change over time.<\/p>\n\n\n\n

    What marketers can do now<\/th>What deserves testing<\/th>What I would avoid<\/th><\/tr><\/thead>
    Communicate product benefits clearly<\/td>FAQ content based on actual consumer questions<\/td>Guarantees of high AI rankings<\/td><\/tr>
    Make product information consistent across websites and retailers<\/td>Comparison pages covering different customer needs<\/td>Conclusions based on one AI response<\/td><\/tr>
    Monitor recommendations across several assistants<\/td>The effect of different product attributes on brand visibility<\/td>Unproven or hidden AI-specific content<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

    None of these actions guarantees the inclusion of a brand in an AI assistant’s answer.<\/p>\n\n\n\n

    They are ways of improving marketing communication and learning about a new discovery channel.<\/p>\n\n\n\n

    7. Why we should be careful when interpreting AI recommendations<\/h2>\n\n\n\n

    Despite these observations, there are several limitations to this experiment.<\/p>\n\n\n\n

    First, the sample size is small. Every brand recommendation probability is based on nine responses for each type of prompt. An additional appearance of one brand increases the probability by approximately 11 percentage points.<\/p>\n\n\n\n

    Second, the comparison included English-language questions about shopping in Montr\u00e9al on free versions of three assistants. The results could be different in another country, another language or for another product category.<\/p>\n\n\n\n

    Third, the experiment was conducted over several days. AI systems and their access to current information can change, and this can cause variations that are not determined only by the wording of the prompt.<\/p>\n\n\n\n

    Fourth, the experiment shows the brands that were included but does not explain why they were selected by the assistants.<\/p>\n\n\n\n

    All those limitations are important since recommendations should not be treated as fixed lists.<\/p>\n\n\n\n

    The 2026 paper by Malthouse et al.<\/a> explored the variability of brand recommendations across different language models. Research carried out by Wharton Generative AI Labs<\/a> demonstrates the dependence of AI shopping recommendations on context. Sielinski (2026)<\/a> also shows why single-run citation visibility measurements can be misleading.<\/p>\n\n\n\n

    It means that AI visibility should be measured carefully. An appearance in one response is not proof of successful positioning, and an absence in another is not proof of marketing failure.<\/p>\n\n\n\n

    Frequently asked questions<\/h2>\n\n\n\n

    How do AI shopping assistants choose which brands to recommend?<\/h3>\n\n\n\n

    It depends on the information available to the system, current sources used during the query and the preferences of consumers. The exact steps differ by AI system, and consumers cannot see all of them.<\/p>\n\n\n\n

    Is GEO going to replace traditional SEO?<\/h3>\n\n\n\n

    No. GEO is based on several aspects that were already important for SEO, such as accessibility, accuracy and credibility of information. That said, marketers now also need to pay attention to the visibility of brands and products in AI-assisted search.<\/p>\n\n\n\n

    Can small brands compete with big ones in AI-assisted recommendations?<\/h3>\n\n\n\n

    They potentially could. In my comparison, Altra became visible when the consumer mentioned wide feet, while not appearing in the general running-shoe answers. It may indicate some opportunities, but additional research is necessary.<\/p>\n\n\n\n

    Do AI assistants recommend the same brands all the time?<\/h3>\n\n\n\n

    No. Even when I asked the same questions, sometimes the order and even the list of brands changed. Multiple observations provide a better basis for comparison than one response.<\/p>\n\n\n\n

    Is there a way for brands to ensure inclusion in AI-assisted recommendations?<\/h3>\n\n\n\n

    There is no guarantee of inclusion in organic recommendations. Better product information and proper positioning may increase brand relevance and visibility, but inclusion is not guaranteed.<\/p>\n\n\n\n

    Conclusion: positioning in the age of AI<\/h2>\n\n\n\n

    Digital marketing has long involved convincing consumers to choose one brand over another.<\/p>\n\n\n\n

    The presence of AI assistants does not diminish that challenge. On the contrary, it creates another scenario where positioning could be helpful: when consumers mention their problems and AI decides which products to recommend.<\/p>\n\n\n\n

    The comparison cannot explain how recommendation algorithms work, but it suggests a clear pattern: well-known brands often appeared in general recommendations, while mentioning a specific need changed the recommendation lists.<\/p>\n\n\n\n

    The examples of Nike and New Balance show why general brand awareness and relevance to consumer needs should not be treated as the same thing.<\/p>\n\n\n\n

    In my opinion, the critical question for marketers is not only whether consumers can find a brand online, but whether that brand is relevant when consumers describe their needs.<\/p>\n\n\n\n

    It does not mean that each company should develop a new strategy for AI. Yet some of the oldest marketing concepts, such as segmentation, differentiation and positioning, could become critical at another stage of the customer journey.<\/p>\n\n\n\n

    A brand should not only check whether AI knows about it. It should ask which consumer problem could lead AI to recommend it.<\/p>\n\n\n\n

    Sources and further reading<\/h2>\n\n\n\n