研究称ChatGPT内84%购买决策源自产品卡片

This study reveals a core phenomenon in AI-assisted shopping: product cards embedded in AI conversations have a decisive influence on final purchase decisions.

1. Key data: 84% of purchase selections came from product cards, and 75% of shopping tasks began with clicking a card. This indicates that users shopping on AI platforms preferentially rely on visual card information rather than searching and comparing on their own.

2. Position drives choice: The first product card was selected 43% of the time, far exceeding random levels, while the second card was selected below random probability. This suggests that "first position" itself functions as a trust signal. Ordinary consumers shopping via AI should be aware that their choices may be guided by ranking mechanisms.

3. Deal cards matter too: When entering the deal-selection stage, 76% of deals pointed to the first card, and the price, stock, and delivery time displayed on the card directly determine the user's next action. Before placing an order, consumers should compare the card information with the actual product detail page to ensure consistency.

4. Practical takeaways: If you shop on AI platforms like ChatGPT, you can proactively ask the AI to change the ranking or request more comparison options, rather than focusing only on the first card. Also, verify that the price and specifications on the card match the final confirmation page to avoid mistaken purchases caused by information mismatch.

This study provides brands with a new operational map for AI shopping scenarios: product cards act as a brand's "virtual storefront" on AI platforms, and competitive decision-making now occurs at the card level rather than on search results pages.

1. Shift in marketing battleground: Within ChatGPT, 84% of final product selections come from in-conversation product cards. Products that fail to appear in the card display sequence lose most purchase comparison opportunities. Brands must therefore make "entering the product card display pool" a core marketing objective.

2. Channel-building priorities: Card ranking position directly affects conversion. The first card captures 43% of selection opportunities, while the second card performs below random probability. Brands should prioritize securing the first display position and strive for the top spot in deal cards as well.

3. Pricing and information consistency: The price, stock, and delivery time shown on cards directly determine the user's next action. Brands must keep product cards, deal cards, and product detail pages consistent in price and specification. The report also documents specific price-mismatch cases, serving as a reminder for price management.

4. Consumer trend insights: User decisions rely heavily on visual card information. Images, titles, and key parameters influence judgment at an early stage. Brands should enhance the intuitiveness of product imagery and parameter presentation, and monitor user responses to sponsored placements to adjust advertising strategies accordingly.

For sellers seeking growth in the AI ecosystem, this study clarifies the new rules of traffic distribution and competition: product cards are the key entry point for AI shopping, and without a card, there is effectively no exposure.

1. Growth market opportunity: ChatGPT and Google AI Mode have become new shopping entry points. Across a study covering 6 categories and 224 shopping tasks, 84% of selections fell on product cards, showing that AI shopping is already a real transaction channel. Early adoption offers a first-mover advantage.

2. Changes in consumer demand: Users no longer make decisions through search lists but instead trust card information embedded in AI responses. The images, prices, and core parameters on cards have become the new basis for purchase decisions.

3. Positive and negative impacts: Securing the first card brings a 43% selection rate, far above random levels. However, the second card attracts selections below random probability. If you cannot secure the top position, the value of exposure drops sharply; merely entering the card pool with a low ranking can become a disadvantage.

4. Actionable measures and takeaways: Focus on two things—pushing products into the product card display pool and making top-position display an operational priority. Also implement the report's 8 card-optimization recommendations, including keeping card information consistent with the detail page to avoid price mismatches that lead to conversion failures.

5. Risk warnings: Not appearing in cards, ranking second, or having inconsistent card information can all lead to rapid exclusion. Sellers need to monitor whether their products appear in AI platform cards and at what rank.

This study offers factories a new production and digitalization orientation based on AI shopping behavior, helping them understand how product data affects sales performance in the AI era.

1. Product parameters and visual design become hard requirements: The images, prices, and core parameters displayed on product cards directly determine whether users make a selection. When designing product details and shooting product photos, factories must consider "card-friendly" presentation, highlighting key specifications to quickly engage users within a small card format.

2. Identifying business opportunities: In AI shopping, the first card's selection rate far exceeds random levels. For factories handling brand orders and OEM work, providing brands with standardized product information packages that are easy for AI to display—high-resolution images, spec sheets, and price tiers—will become more competitive.

3. Digitalization and e-commerce implications: Factory digitalization should not stop at production. It must extend to seamless connectivity from production data to AI shopping cards. Master data (price, spec, stock) must remain consistent across systems to avoid the price-mismatch cases documented in the study, since AI platforms directly scrape this information to generate cards.

4. Product R&D considerations: The study shows users rely heavily on card information when making purchase decisions. In the new-product planning stage, factories should treat "easy for AI to understand and recommend" as part of product design, providing clear, quantifiable selling points to reduce friction for brands later when they perform AI optimization.

This study serves as an important reference for the AI shopping services industry. It identifies new client pain points and offers service providers practical, packagable business directions.

1. Industry trend: AI assistants are becoming the entry point for shopping decisions, and product cards are the "brand storefront." In the future, more e-commerce transactions will occur within conversational interfaces. Service providers should monitor this trend and build AI shopping behavior optimization services early.

2. New technology applications: The study captured users' full-funnel behavior in ChatGPT and Google AI Mode, revealing a strong correlation between card display position and selection probability. This provides a data foundation for service providers to develop ranking monitoring and card exposure diagnosis tools.

3. Client pain points: Existing AEO/GEO optimization by brands and retailers only addresses exposure. True conversion competition happens at the card and deal-card stage. Clients do not know whether their products are in the card pool, at what rank, or whether card information matches the detail page. These gaps are precisely where service providers can offer consulting.

4. Solution approach: The ReFiBuy team has translated the research findings into 8 actionable optimization recommendations. Service providers can directly adopt these directions to offer clients "agentic commerce optimization," including pushing products into the display pool, securing top rankings, and maintaining card information consistency. They can also develop audit and remediation methods based on the price-mismatch cases.

This study provides empirical evidence for platform operations and merchant recruitment in AI shopping interfaces, highlighting the significant impact of product card ranking mechanisms on transaction conversion.

1. Platform's commercial demand: Product and brand parties have a strong need for "card display positions" and will begin competing for the top spot. This creates opportunities for platforms to design paid displays or sponsored cards, while also requiring more transparent and clearer ranking rules to avoid disputes.

2. Reference for latest platform practices: The report examines user reactions to sponsored placement positions and suggests that platforms plan card and sponsored layouts reasonably within AI conversation interfaces, balancing user trust with commercialization space.

3. Operational management insights: Card ranking greatly influences the probability of product selection—the first card gets 43% while the second remains below random probability. Platforms should examine whether existing ranking algorithms favor the first position and whether this creates user selection bias. If necessary, randomization or diversified display may be introduced.

4. Risk avoidance and optimization: The study documents cases where users encountered product card price mismatches during the decision process. This reminds platforms to strengthen review of card information, ensuring consistency between cards and detail pages. Otherwise, price misinformation may cause user churn and compliance risks. Meanwhile, platforms should enhance the structuring of card images, prices, and parameters to help merchants accurately enter the display pool.

This study provides new empirical evidence on AI shopping behavior, identifies emerging industry trends, and raises governance issues worthy of deeper investigation.

1. New industry dynamics: Product cards embedded in AI assistants have become the core interface for purchase decisions. With 84% of final selections coming from cards, e-commerce traffic entries are shifting from search lists to conversational recommendations, forming a new retail infrastructure.

2. New problems and decision bias: The study reveals a clear primacy effect—the first card's selection rate of 43% is far above random, while the second card falls below random probability. This ranking bias may lead to user choices being manipulated by algorithms, raising concerns about consumer autonomy and fair competition that merit further research.

3. Policy and regulatory implications: AI platforms should clearly label sponsored or advertising information when displaying product cards and deal cards, avoiding the use of positional advantage to mislead consumers. It is advisable to introduce transparent oversight of card information authenticity, consistency, and ranking rules. Price mismatches in particular may violate advertising law and consumer protection law.

4. Business model research: The study introduces the concept of "Agentic Commerce Optimization" (AIBO), which focuses on operational optimization around product cards and deal cards. This is a new business model based on AI shopping scenarios and can create value for brands and service providers. Additionally, comparing product cards to the "golden shopping cart" offers a fresh perspective on e-commerce platform economics.