Take Caution in Using LLMs as Human Surrogates: Scylla Ex Machina
Large language models (LLMs) are used in consumer research, but how closely do they simulate human behavior? Eight popular LLMs tested fail to reproduce decisions humans make in a simple game of strategy. In their current state, LLMs are better used as supplementary tools rather than substitutes for conventional research.
Author: Yuan Gao, Dokyun Lee, Gordon Burtch and Sina Fazelpour

Can Large Language Models Extract Customer Needs as Well as Professional Analysts?
Extracting consumer needs from online reviews and other sources can inform product and service innovation but is typically costly. Large language models (LLMs) trained to professional standards can cost-effectively identify implicit consumer needs that are specific but abstract enough to be useful. Automating other innovation tasks may be more difficult.
Author: Artem Timoshenko, Chengfeng Mao, and John R. Hauser

Revolutionizing Marketing Research with a Large Language Model: A Hybrid AI-Human Approach
The demand for “faster, better, cheaper” once presented marketing researchers with an unsolvable challenge. Today generative AI large language models can assist at many stages of qualitative and quantitative research but keeping humans in the loop will still be critical to ensure the best results and guard against inadvertent biases.
Author: Neeraj Arora, Ishita Chakraborty and Yohei Nishimura

AI-Powered Digital Streamers in Online Retail: Empirical Insights and Design Strategies from Experiments
As influencers and social shopping sites proliferate, retailers need scalable ways to engage consumers. AI-assisted digital streamers can help, but retailers must focus on features that replicate human engagement. Behavioral realism—real-time Q&A and other interactive features—are among the most effective strategies making AI streamers competitive with human influencers.
Author: Yahui Liu, Lei Wang, Shuai Yang and Yanwen Wang

Dynamic Personalization with Multiple Customer Signals: Multi-Response State Representation in Reinforcement Learning
Reinforcement machine learning can help companies understand their customers, but first-party data is typically noisy and focused on short-term effects that may not optimize long-term outcomes such as customer lifetime value (CLV). A novel Multi-Response State Representation (MRSR) Learning method can address these challenges to go beyond next best actions.
Author: Liangzong Ma, Ta-Wei Huang, Eva Ascarza and Ayelet Israeli

Get Rid of It! How Interface Layouts Influence Product Retention Behaviors
In today’s “always on” social media world, continuously scrolling content can offer an engaging experience but may not be ideal for task-oriented shopping. To avoid wasteful product disposal or expensive returns, online retailers should use discrete pagination or product numbering on continuous websites to enhance consumers’ control and purchase confidence.
Author: Huitian Zhang, Lei Su and Jaideep Sengupta

Stream-In or Stream-Out? Comparing Sales Impacts of In-House and Influencer-Partnered Livestreaming
Today the “make or buy” decision facing brands is whether to partner with influencers or do their own livestreaming? While “mega” influencers can help build brand awareness, goodwill, and reputation over time, in-house livestreaming is a cost-effective approach for smaller brands to increase sales, improve visibility, and foster customer loyalty.
Author: Zining Wang, Yanwen Wang, Shuai Yang and Hongju Liu

What Makes Players Pay? An Empirical Investigation of In-Game Lotteries
Purchasing assets from “loot boxes” can help players advance levels in free-to-play online games. However, the uncertainty associated with this may constitute a form of gambling for those few who become major sources of revenue. Simulations revealing players’ divergent preferences and motivations can suggest whether and what regulation is appropriate.
Author: Tomomichi Amano and Andrey Simonov

How AI Chat Agents’ Conversational Style Influence Customers
Where and how AI agents can effectively automate service encounters and other customer experience may depend on the type of conversational content involved. Customers may accept AI agents’ use of informational and normative content (conversational politeness cues) but resent attempts at relational content (“how are you?”) perceived to be inauthentic.
Author: Feyzan Karabulut, Sarah G. Moore and Paul R. Messinger

Does behavioral spillover follow policy changes? A longitudinal perspective on changing plastic regulation
Plastic bag regulations can have positive spillover effects such as reduced purchase of plastic bottled water. Increasing the stringency of regulations, however, may create at least a temporary backlash from consumers. Policymakers must consider—and manufacturers and retailers anticipate—how sequential changes in regulations will affect consumer preferences and behavior.
Author: Mengfei Ye and Jenny van Doorn

The Perceived Causality in Benefit/Cost and Cost/Benefit Ratios
Metrics used to demonstrate marketing effectiveness can be mathematically equivalent but have opposite effects. Revenue/cost ratios such as Return on Advertising Spend are more likely than cost/revenue ratios such as Advertising Cost of Sales to persuade decision-makers that advertising causes sales but may also reduce their willingness to risk innovations.
Author: Archer Pan, Jean-Louis Sterckx, Bart De Langhe and Stijn van Osselaer

When Food Gets Political: How Does News Drive Green Food Offerings?
Start-up companies in new categories often have limited marketing budgets. Free local news coverage and social media posts can help local merchants to adopt innovative products, especially where they align with consumers’ views on environmental sustainability or other values, achieving results comparable to traditional TV advertising of consumer package goods.
Author: Boya Xu, Tong Guo and Daniel Yi Xu

Words That Matter: Analyzing the Causal Effect of Words
How you say it makes all the difference in marketing (and life). Realistic A/B tests that vary multiple phrases can’t determine the causal effects of specific words. An alternative approach can control possible confounds of wording and context, yielding true causal insights that can be used to optimize marketing content.
Author: Alain Lemaire, Mingzhang Yin and Oded Netzer

Balancing Engagement and Polarization: Multi-Objective Alignment of News Content Using LLMs
Media and marketers need content that attracts audiences while maintaining editorial or brand positioning. Generative AI can produce content that is engaging for some but may alienate others. To mitigate this polarization risk, existing LLMs can and should be adapted to balance multiple competing objectives such as engagement and alignment.
Author: Mengjie (Magie) Cheng, Elie Ofek and Hema Yoganarasimhan

Used Enough? The Effect of Categorization on Product Replacement Timing
Replacing products too early creates unnecessary waste; waiting risks consumers’ health and safety. Displaying a product’s lifespan with more rather than fewer stages signals that it is no longer safe and useful. This retrospective perception of “usage sufficiency” encourages timely replacement more than does the prospect of insufficient remaining value.
Author: Poornima Vinoo, Grant E. Donnelly, Mathew S. Isaac and Aaron R. Brough

Platform Recommendation Algorithms, Niche Market Entry, and Quality Competition
Prohibiting self-preferencing by an online retailer’s recommendation algorithm aims to protect consumers but may have unintended consequences depending on how sellers and buyers respond. The UK’s regulation of Amazon’s algorithm did promote price competition, but primarily in lower-rated niche markets while reducing overall variety, consumer purchases and average product ratings.
Author: Gaoyang Cai, Xia Han and Grace Gu

Market Structure Mapping with Interpretable Visual Characteristics
Consumers choose autos and other products on both functional and visual attributes. An advanced machine learning technique can help marketers map competitive markets on both dimensions simultaneously, identifying product boundaries and potential extensions, aligning with consumer decision processes and assessing competitive threats better than maps based only on functional attributes.
Author: Ankit Sisodia, Vineet Kumar

Beyond the Spike: How Stakeholder Response Patterns Predict Retail Crises and the Impact of Retailer Reactions
Controversies about product failures or corporate policies can affect investors. Such brand crises tend to follow typical trajectories of media coverage and public reaction. The effectiveness of a company’s response will depend on identifying the specific trajectory and the timing and content rather than the number of its press releases.
Author: Koen Pauwels, Kelly Hewett, Raoul Kuebler and Meike Eilert

Generative AI as a Research Confederate: The LUCID Methodological Framework and Toolkit for Human-AI Interactions Research
As companies increasingly rely on AI chatbots to interact with customers, it is essential that the information they provide be relevant and accurate and delivered in an appropriate manner.  Prompting LLMs to act as research confederates can help design experiments managers can use to safely pre-test chatbots before deploying them.
Author: Aaron M. Garvey and Simon J. Blanchard

A Helpful Tool or Marketing Gimmick? How Using Augmented Reality Pre-Purchase Shapes Consumer Response to Product Failure
Augmented Reality (AR) allows online shoppers to visualize a product in a scanned real setting, potentially improving choices and reducing returns. When the delivered product fails to meet expectations, however, shoppers may infer AR was used to deceive rather than inform them, blame the brand, and reduce their future patronage.
Author: Jianna Jin and John P. Costello

A Multiple-Stakeholder View of Open and User Innovation: Systematic Review and Relational Synthesis
Innovation involves more than creators. This extensive review of research uses stakeholder theory to offer a framework of the distinct roles of creators, contributors, and customers in open and user innovation (OUI) to categorize and evaluate the full spectrum of innovation participants and, more importantly, identify who has been overlooked.
Author: Keith Marion Smith, Matthew S. O’Hern, Mason R. Jenkins, Paul W. Fombelle and Charles H. Noble

Marketing Employees with AI Expertise: Roles and Performance Implications
How firms manage marketing AI human capital sends a message to Wall Street. Attracting such talent boosts stock value but losing it hurts even more, especially junior staff potential change agents. Larger firms can leverage greater organizational resources but should adapt to the needs of diverse markets and business units.
Author: Suyun Mah, Chengxin Cao, Eunyoung Song and Ju-Yeon Lee

Marketing the Future: How Deep Uncertainty Shapes the Future of Marketing
Most research on decision-making assumes some knowledge of the time-discounted value and probabilities of specific outcomes. Deep uncertainty arising from both “unknown unknowns” and complex systems of “known unknowns” challenges such models. This conceptual framework for future research can help companies understand consumer decision-making and resource allocation under deep uncertainty.
Author: Stefan Stremersch, Nuno Camacho, Benedict G.C. Dellaert, Eric J. Johnson and Roland T. Rust

Customer Privacy Journey
The data companies need to offer seamless customer experiences can raise privacy concerns at several points in the journey. Privacy fatigue from past data breaches can fester despite apparently renewed customer satisfaction. Companies need to move beyond reactive recovery from data breaches to proactively managing privacy as an ongoing relationship.
Author: Natalie Chisam, Jordan W. Moffett and Kelly D. Martin

Building Persuasive Stories with Emotion Sequences
Emotional stories can motivate charitable donations, but the sequence of specific emotions determines their effectiveness. Audiences are most engaged by stories that move from sadness to caring and allow them to identify with the protagonist. Stories rewritten by AI-assisted professionals or by trained AI alone can be even more effective.
Author: Samsun Knight, Liu Liu, and Laura J. Kornish

In Privacy We Trust: The Effect of Privacy Regulations on Data Sharing Behavior
Companies can offer tangible rewards for purchase and other personal data—but what really motivates consumers to share? Aligning with data privacy regulations that require transparency and allow opt out may be even more effective, creating trust, engagement and sharing by reassuring consumers that their data will be used appropriately.
Author: Ozge Demirci, Ayelet Israeli and Eva Ascarza

Out with the New, In with the Old: The Impact of Incremental Innovations on Market Share Gains
Incremental innovation that creates cost-effective, tangible customer value is the key to growing market share, especially for non-dominant companies in mature R&D intensive business sectors. Rather than relying on brand or R&D spend alone, keeping close to customers can help identify informed buyers and influencers who will act as advocates.
Author: Rodrigo Farinha, Leandro Guissoni, Jonny Rodrigues and Thales Teixeira

Generative AI and the Commoditization of Marketing Knowledge
Busy managers may lack ready access to foundational marketing knowledge established by rigorous academic research, relying instead on intuition. Generative AI in the form of Large Language Models (LLMs) can help close this gap, providing frictionless access to expertise that levels the playing field for all firms, especially smaller start-ups.
Author: Raymond R. Burke, Maximilian Matthe and Alex Leykin

From Prompt to Product: Reimagining Visual Search with Generative AI
GenAI visualizations can improve search in online retail, helping consumers articulate their preferences, providing feedback that they have been understood, and increasing their satisfaction and purchase intention. Visualizations that accurately reflect consumer specification for the design can enhance the experience even when the match to available products is not perfect.
Author: Jan Ole Krugmann

Generative AI in Equilibrium: Evidence from a Creative Goods Marketplace
GenAI images crowd out others in creative goods markets, increasing overall quality and variety for consumers but harming creators by reducing the value of original content used to train models. Copyright protections ensuring “fair use” not only compensate non-GenAI producers but can also mitigate crowd-out by increasing GenAI production costs.
Author: Samuel G. Goldberg and H. Tai Lam

Who Expands the Human Creative Frontier with Generative AI: Hiveminds or Masterminds?
Text-to-image AI models diffused from “mastermind” early adopters to the “hivemind” can be a catalyst for human creative expression by facilitating increased overall output.  Without this productivity effect, however, AI-assisted creators seem to experience a dilution effect when leveraging AI tools where on average their creativity regresses to the mean.
Author: Eric B. Zhou, Dokyun Lee and Bin Gu

Stop the Spread: A Bi-Partisan Approach to Aligning Warning Content with Consumer Segments to Mitigate the Spread of Misinformation
Misinformation can reflect and exacerbate political and other differences. Even prosocial sharing to warn others can spread misinformation. Ideally, warning labels should be tailored to the beliefs, values, and identities of the audience they target. Women respond better than men, but where targeting is not feasible, generic warnings may serve.
Author: Marina Cozac, Martin Mende, Maura L. Scott, Christopher Berry, and Beth Vallen

A Qualitative Analysis of Consumer Barriers to Mental Health Consumption
Stigma may deter some from seeking mental healthcare, but low motivation to initiate care, search friction finding appropriate providers, skepticism about provider motives and efficacy, and constraints of cost, time, and availability loom larger. Rather than reducing stigma to boost mental health consumption, boosting mental health consumption may reduce stigma.Author: Justin Pomerance

Spatial Marketing Research: Leveraging 3D Virtual and Interactive Spaces to Study Marketing Phenomena
Spatial technologies such as the metaverse and immersive virtual reality can accelerate innovation and inform marketing, but the relevant literature is widely dispersed. Grounded in activity theory and design research, this Spatial Marketing Research framework can help managers understand how these technologies can be implemented to create real business value.
Author: Sebastian Hohenberg, Thorsten Hennig-Thurau, Ronny Behrens, Patrick Wöhnl and Hanna Pott

Leveraging Large-Scale Granular Single-Source Data for TV Advertising
Causal models of ROI must control endogeneity resulting from ads targeted at specific audiences and times. True experiments are difficult and expensive, but the random assignment of ad slots within linear TV shows can yield the required instrumental variable. Otherwise, observational models may greatly overstate the impact of TV advertising.
Author: Tsung-Yiou Hsieh, Rex Yuxing Du and Shijie Lu

Predicting Behaviors with Large Language Model (LLM)-Powered Digital Twins of Consumers
Large language models (LLMs) can be fine-tuned on structured data from surveys and CRM transactions and unstructured data from product reviews and social media posts. Retrieval-augmented generation (RAG) can add contextual data such as category, brand and product knowledge. The resulting “digital twins” accurately predict both consumers’ choices and evaluations.
Author: Bingqing Li, Qiuhong (Owen) Wei and Xin (Shane) Wang

Designing with Edge Consumers: How Inclusive Design Orientation Transforms New Product Development
New product development typically targets larger, core consumer segments, but the alternative “life hacks” used by excluded consumers can be a rich source of innovative ideas. Internal champions who promote an inclusive design orientation (IDO) embed “day-one inclusion” in all NPD processes, enhancing innovation, revenue, employee engagement and brand loyalty.
Author: Vanessa Patrick, Deepa Chandrasekaran, BJ Allen

Using LLMs for Market Research
Baseline LLMs can simulate consumers’ WTP for existing features in a specific product category but need fine-tuning with prior human results to accurately simulate response to new features in that category. Simulating demographic differences may require separate models. Fine-tuning may not improve LLMs ability to extrapolate to unrelated product categories.
Author: James Brand, Ayelet Israeli, Donald Ngwe

Fair Document Valuation in LLM Summaries via Shapley Values
LLM summaries of search results may enhance user experience and platform engagement but also reduce visits to sites that compensate content providers. An algorithm that efficiently ensures fairness, generality and scalability can be applied by search, review and other platforms to align the incentives of content creators and end users.
Author: Zikun Ye and Hema Yoganarasimhan

Learning When to Quit in Sales Conversations
Outbound sales reps must decide whether to persist with a given call to convert or quit and spend time on another, higher-potential call. LLM-based stopping agents trained on transcripts alone can detect linguistic cues predicting failure sooner than reps themselves, helping them shift to more productive uses of their time.
Author: Emaad Ahmed Manzoor, Eva Ascarza, Oded Netzer

Characterizing and Minimizing Divergent Delivery in Meta Advertising Experiments
Targeting algorithms can match specific content and audience segments. In the case of A/B tests, this “divergent delivery” can make it difficult to distinguish the effects of creative content from the targeting itself. Careful campaign configuration can largely mitigate this, allowing meaningful creative comparisons while not disrupting standard marketing practices.
Author: Gordon Burtch, Robert Moakler, Brett R. Gordon, Poppy Zhang and Shawndra Hill

Estimating Treatment Effects under Recommender Interference: A Structured Neural Networks Approach
Content-sharing platforms use A/B tests to improve algorithms matching creators and audiences. Selection bias can result from highly personalized recommendation feeds or when tests boost treated content. This cost-effective alternative combines a structural choice framework with neural networks to account for rich viewer-content heterogeneity, improving results and business decisions.
Author: Ruohan Zhan, Shichao Han, Yuchen Hu and Zhenling Jiang

Fragile Futures: How Consumers Respond to Future-Oriented Interventions
Consumers have many reasons and pathways for adopting—or rejecting—new technologies such as electric vehicles. Like customer journey mapping, the theoretical lens of future-making can help marketers and policymakers understand how different consumers evaluate, negotiate and enact alternative futures that align their goals with desired business and policy outcomes.
Author: Daiane Scaraboto, Alison M. Joubert, Claudia Gonzalez-Arcos and Jorgen Sandberg

Experimentation in Online Advertising
Randomized controlled trials (RCTs) allow causal inference about digital ad effectiveness but incur both direct and opportunity costs by excluding control groups. Online platforms vary costs for ads and for support of experimentation to attract bidders. This game-theoretic model can help both parties determine the costs and benefits of experimentation.
Author: W. Jason Choi and Amin Sayedi

The Business Case for Selling and Marketing to Singles
As more individuals choose to remain single, marketers must rethink products, pricing, policies, and brand messaging based on two-adult households with pooled income, shared responsibilities, and synchronized consumption. Firms designing for families of one, rather than retrofitting couple-centric offerings, can gain revenue and long-term loyalty from a rapidly expanding market.
Author: A. Peter McGraw

Consumer Responses to Premium Services: The Role of Zero-Sum Beliefs
The Pareto rule (20% of customers account for 80% of revenue) tempts companies to focus on premium services. However, perceiving these offerings as zero-sum and thus unfair can diminish the appeal of basic offerings for non-premium customers. Companies can mitigate backlash by promoting premium offerings that subsidize services for all.
Author: Serena Haggerty, Christilene du Plessi and Debora Viana Thompson

AI in Disguise – Quasi-Experimental Analysis of a Large-Scale Deployment of AI-Generated Display Ads
GenAI can produce the specific image features that make ads undistinguishable from “sibling” human-made ads—which they outperform on measures like click-through rates. Advertisers can use a “looks-like-AI” metric to predict and mitigate artificiality penalties before deployment. Platforms can accelerate adoption by targeting performance-oriented advertisers most receptive to genAI technologies.
Author: Yannick Exner, Jochen Hartmann, Oded Netzer, Shunyuan Zhang and Ziqian Ding

When Connections Depress Contributions: The Hidden Cost of Social Filtering
Recommendation algorithms can help consumers find useful information. Prioritizing content from a user’s social connections can strengthen engagement on online platforms, but such social filtering may also erode the volume, efficiency, quality and diversity of user-generated content. Platforms need to carefully balance homophily-driven recommendations with mechanisms that promote diverse interactions.
Author: Ziwei Cong, Jia Liu and Shijie Han

Modeling Aggregate Consumer Journeys in a Cookie-Free Environment: A Two-Stage Framework for Marketing Mix Models
Marketing mix models developed for the era of TV, radio and print advertising struggle in today’s complex media environment, compounded by the loss of individual-level data. A two-stage approach using machine learning classification to predict directional changes followed by classical time-series techniques can extract meaningful patterns from noisy, aggregated data.Author: Victor Churchill, H. Alice Li and Dongbin Xiu

Are Display Ad Content and Landing Page Ad Content Complements or Substitutes? A Field Experiment
The effectiveness of B2B display and landing page ads in customer acquisition depends on prior knowledge and the content and sequence of the ads. In low-information markets, repeated content on product quality drives engagement. In high-information markets, display ads signal potential matches with customer preferences, reinforced by the landing-page ad.
Author: Yewon Kim and Kirthi Kalyanam

The Brand Backstory and Strategic Performance of Transparency
Brands may offer “inside access” such as factory tours or visitor centers to build trust, differentiation and engagement, but these benefits require carefully curated transparency to shape what consumers see as “authentic.” More than decorative experiences, backstory performances must provide consumers with genuine insights into company values, processes, or heritage.
Author: Cristel Antonia Russell, Anne Hamby, Stephanie Feiereisen and Hope Jensen Schau

Why it Works: Can LLM Hypotheses Improve AI Generated Marketing Content?
A/B tests to optimize marketing content can determine what works—but not why. This limits generalization and may even optimize for “click bait” rather than truly engaging content. Using LLMs to first generate and test hypotheses about audience reactions can help marketers generate more effective, trustworthy, and strategically aligned content.

Author: Tong Wang, K. Sudhir and Hengguang Zhou

“No-Interest” Advertising in High-Interest Loan Market
Cash advance apps may be useful, but consumers can be misled by complex fee structures and ads touting “no interest” loans. Experienced users may be even more susceptible, and making upfront costs more salient doesn’t fully mitigate the impact of ads. Regulation may be needed to restrict lenders’ misleading advertising.
Author: Yufeng Huang, Yukun Liu, Bowen Luo and Xi Wu

The Intended and Unintended Consequences of Privacy Regulation for Consumer Marketing: A Marketing Science Institute Report

• Jean-Pierre Dubé, University of Chicago
• Dirk Bergemann,Yale University
• Mert Demirer, Massachusetts Institute of Technology
• Avi Goldfarb, University of Toronto
• Garrett Johnson, Boston University
• Anja Lambrecht, London Business School
• Tesary Lin, Boston University
• Anna Tuchman, Northwestern University
• Catherine Tucker, Massachusetts Institute of Technology
• John G. Lynch, University of Colorado-Boulder & Marketing Science Institute

 

The Marketing Uncertainty Whitepaper: What to Avoid, What to Do Now, and What to Build for the Future

• Keith Smith, Marketing Science Institute
• Earl Taylor, Marketing Science Institute