Are AI Chatbots Changing How We Shop?
For decades, online search was the starting point for many of our purchases. Then, in just a few years, ChatGPT and other AI tools burst onto our computer and phone screens. What does shopping with an AI assistant change for consumers—and for the sellers and advertisers trying to reach them? We asked Yale SOM economist Jidong Zhou, whose research interests include consumer search and information design.
Do you think AI has the potential to dramatically change the model for consumer search?
Compared with traditional search engines (e.g., non-AI Google search), AI-assisted search tools such as ChatGPT offer several advantages, especially when consumers are searching for complex and personalized products, such as vacation packages or wellness programs. First, instead of relying on simple keywords, consumers can now describe their needs in greater detail and with higher precision, leading to more targeted and relevant results. Second, if a consumer has interacted with, say, ChatGPT over time, the system learns a great deal about the user—often more comprehensively than Google—which further refines the search outcomes. Third, AI-assisted search typically provides detailed summaries of each option and can even generate comparison tables. As a result, consumers may no longer need to click through individual links to check details themselves—a key difference from traditional search that further reduces search costs.
Should we call these changes dramatic? I’m not sure. From a search theory perspective, they primarily make searches more informative and less costly. Personalized search results from Google or Amazon had qualitatively similar effects in the past, though this time the magnitude of the impact is likely to be much bigger.
As consumers begin to receive more personalized and context-specific product suggestions, firms must also rethink their pricing, differentiation, and product design.
The landscape of consumer search and shopping, however, could change more dramatically as firms and AI companies deepen their integration. For example, Walmart and OpenAI recently announced a partnership that will soon enable Walmart customers to shop through ChatGPT using its “instant checkout” feature. Consumers can simply chat with the AI to describe what they want to buy—or even outline what they plan to do (for example, hosting a party with specific requirements)—and the AI will find the relevant products on Walmart’s platform and prepare them for checkout.
Significant changes may also occur as more consumers begin using web browsers with deeply embedded AI tools, such as ChatGPT Atlas, launched by OpenAI a few weeks ago. For instance, when a consumer browses a shopping website using Atlas, ChatGPT appears in a sidebar and can instantly help analyze and compare options or answer questions. It’s almost like having a personalized and highly knowledgeable assistant shopping alongside the consumer. Similarly, when a user reads a recipe on a cooking website, ChatGPT can immediately locate where to buy the ingredients.
What are we learning about how consumers are using AI so far?
According to a recent survey by Deloitte, 53% of surveyed consumers are experimenting with generative AI or using it regularly, up from 38% in 2024. People use AI for various activities such as writing, research, advice, content creation, etc. However, very little rigorous research based on real data has been done on how consumers use AI.
A very recent study on this topic was conducted by a group of economists in collaboration with OpenAI’s Economic Research team. According to this paper, most ChatGPT usage falls into three categories, which the authors call practical guidance, seeking information, and writing. Notably, the share of messages classified as seeking information rose from 18% in July 2024 to 24% in June 2025, highlighting the ongoing shift from traditional web search toward AI-assisted search.
How should companies prepare for a world in which consumers are getting product information from AI?
AI chatbots such as ChatGPT have not yet adopted a pay-for-recommendation or pay-for-ranking business model, but this is likely to change in the near future. At this stage, companies should understand better how AI chatbots gather and summarize product information, and how they rank competing options. To ensure accurate and favorable representation, firms should make their websites more AI-readable—using structured data, clear and verifiable product descriptions, and transparent performance metrics. In the future, as AI platforms adopt business models similar to those of Google or Meta, companies will need to balance visibility strategies—whether through direct integration, partnerships, or potential paid placements—with maintaining credibility and consumer trust.
On the other hand, as consumers begin to receive more personalized and context-specific product suggestions, firms must also rethink their pricing, differentiation, and product design. For instance, when more consumers are shown their preferred products first, firms effectively gain additional market power, which may lead them to raise prices—partially offsetting the consumer benefits from more efficient search. (This is a point I made in my own research papers “Improved Information in Search Markets” and “Personalization and Privacy Choice.”) At the same time, as personalized matching grows in importance, it is also desirable for firms to expand their product varieties to better cater to heterogeneous consumer preferences.