How ChatGPT product results are selected, displayed and ranked

OpenAI unveiled a major update to ChatGPT five days ago on April 28, 2025, introducing product recommendation capabilities that fundamentally differ from conventional search engines. The system, which has already processed over 1 billion web searches in the past week, employs a distinctive methodology for selecting, displaying, and ranking products that marketers and consumers should understand.

The new shopping functionality, now rolling out to Plus, Pro, Free, and logged-out users globally, operates without sponsored placements. Instead, it uses a sophisticated algorithm to recommend products based on user intent and preferences.

The product selection algorithm

ChatGPT determines which products to display through an intent-based evaluation process. When a user asks about products (for example, dog costumes), ChatGPT activates its shopping features, displaying what OpenAI describes as "visually rich carousels" with product information and purchase links.

Unlike Google or Amazon, ChatGPT's product results "are selected by ChatGPT independently and are not ads," according to OpenAI's documentation. This represents a significant departure from the advertising-driven approach of traditional e-commerce platforms and search engines.

Four key factors in product selection

When determining which products to display, ChatGPT employs a sophisticated decision framework that incorporates multiple data sources and contextual elements. According to OpenAI's documentation, the system weighs four primary factors:

1. Structured metadata from third-party providers - This includes technical specifications, pricing information, and product descriptions
2. Third-party content - Consumer reviews and ratings that help gauge product quality and satisfaction
3. Pre-search model responses - Information generated by ChatGPT before incorporating new search results
4. OpenAI safety standards - Filtering mechanisms that ensure recommended products meet company guidelines

The relative importance of these factors shifts dynamically based on user requirements. For example, when a user specifies a budget constraint, the system prioritizes price data over other considerations. Similarly, when specific product attributes are mentioned, those characteristics receive greater weight in the selection algorithm.

How user context influences product selection

ChatGPT's product recommendations are heavily influenced by contextual signals derived from user interactions. The system processes:

1. Explicit query information - Direct product requirements stated in the search
2. Memory data - If enabled, ChatGPT references previous conversations to personalize recommendations
3. Custom instructions - User-defined preferences that persist across sessions

This contextual awareness creates a personalized shopping experience. For instance, if a user had previously indicated certain preferences or aversions, ChatGPT integrates this information into its product selection logic. According to OpenAI's documentation, if a user "had previously indicated a dislike for clowns, the model might also consider that and leave out clown costumes" from dog costume recommendations.

Users maintain the ability to refine these recommendations through follow-up questions and clarifications, allowing for an iterative selection process.

Product information display techniques

OpenAI has implemented several techniques to optimize how product information appears to users:

1. Standardized product descriptions - ChatGPT generates simplified, consistent product titles and descriptions that normalize the inconsistent terminology used by different merchants for identical products
2. AI-generated feature labels - Products may display designations like "Budget-friendly" or "Most popular" based on ChatGPT's analysis of available data, though these labels "are not guarantees or verified statements"
3. Review synthesis - The system creates concise summaries that highlight common user opinions about products, drawing from reviews found on public websites
4. Rating aggregation - Star ratings shown in the interface may represent aggregated scores from multiple sources, which "may not match the rating available on any particular website"

Merchant ranking methodology

The current merchant selection system functions primarily as a presentation layer rather than a sophisticated ranking engine. When users click on a product, ChatGPT may display a list of merchants offering that item, with placement "predominantly determined by these [third-party] providers" rather than by performance metrics like price competitiveness or shipping policies.

Price information flows from these same third-party providers. After clicking on an initial price (which typically reflects the first merchant listed and "may not be the lowest available price"), users can view additional pricing options from other merchants offering the same product.

OpenAI acknowledges certain limitations in this approach. Price updates from merchants may experience delays before appearing in ChatGPT, and estimated taxes and delivery fees might differ from final charges. The company indicates it is "actively working on faster methods to update this information."

Technical underpinnings of product selection

The technical implementation of ChatGPT's product recommendation system reveals several important aspects of how selections are made and presented:

Query rewriting for product search

When users express shopping intent, ChatGPT employs a sophisticated query rewriting mechanism to enhance search precision. Rather than using the exact language provided by users, the system transforms natural language queries into structured search parameters.

For example, when a user asks "I'm looking to buy costumes for my two dogs," ChatGPT might internally reformulate this into a more technical query focusing on product attributes, sizes, and categories. This reformulation process helps bridge the gap between conversational language and the structured data needed for effective product retrieval.

Location-awareness in product selection

Product relevance is further enhanced through location data processing. The system "collects general location information based on your IP address and may share that general location with third-party search providers to improve the accuracy of your results," according to OpenAI's documentation.

This location awareness enables ChatGPT to deliver regionally appropriate product recommendations. For instance, when a user asks about nearby stores or region-specific products, ChatGPT might detect their location as San Francisco and adjust recommendations accordingly, without sharing the actual IP address with third parties.

Roadmap for merchant integration

OpenAI has indicated plans to develop more direct relationships with merchants. The company is "exploring ways for merchants to provide us their product feeds directly, which will help ensure more accurate and current listings," suggesting a future where product selection may rely less on third-party aggregators and more on direct merchant data.

This approach would potentially address current limitations in price and inventory accuracy, while giving merchants more control over how their products appear in ChatGPT's recommendations.

How memory and custom instructions enhance product selection

Two auxiliary systems significantly influence ChatGPT's product selection algorithm, creating a more personalized shopping experience:

Memory integration in product queries

When the Memory feature is enabled, ChatGPT leverages previously established user preferences to refine product recommendations. According to OpenAI's documentation, "If Memory is enabled, when ChatGPT search rewrites your prompt into a search query it may also leverage relevant information from memories to make the query better and more useful."

This creates powerful personalization capabilities. For example, if ChatGPT has recorded that a user is vegan and lives in San Francisco, a query for "restaurants near me that I'd like" might be internally rewritten as "good vegan restaurants San Francisco," delivering highly personalized results without requiring the user to specify these preferences repeatedly.

Custom instructions as persistent preferences

The Custom Instructions feature serves as another important input to the product selection algorithm. These instructions allow users to establish persistent preferences that influence all future interactions, including shopping queries.

OpenAI notes that custom instructions "allow you to share anything you'd like ChatGPT to consider in its response." In shopping contexts, this might include budget constraints, style preferences, or ethical considerations that consistently shape product recommendations across multiple sessions.

Importantly, these instructions "will be added to new conversations going forward" and can be edited or deleted at any time, giving users granular control over the factors influencing their product recommendations.

Marketing implications of ChatGPT's product selection approach

ChatGPT's product selection methodology creates five critical implications for marketing professionals:

1. Non-advertising product placement

Unlike traditional e-commerce platforms where product visibility can be purchased through advertising, ChatGPT's approach relies on algorithmic selection without sponsored placements. OpenAI explicitly states that products "are selected by ChatGPT independently and are not ads." This fundamental difference requires brands to focus on organic factors rather than paid placement.

2. Review-centric optimization

Product reviews have outsized importance in ChatGPT's selection algorithm. The system synthesizes reviews from public websites and frequently uses this information for both ranking and feature labeling (e.g., identifying products as "Budget-friendly" based on review mentions of value). This elevates the strategic importance of generating authentic positive reviews rather than simply optimizing product titles and descriptions.

3. Data accuracy requirements

Since ChatGPT generates "simplified product titles and descriptions based on information from third-party providers," the accuracy of upstream product data becomes critically important. Inconsistent or outdated information in third-party databases could result in products being misrepresented or excluded from recommendations entirely.

4. Visual merchandising transformation

The "visually rich carousels" ChatGPT employs for product display create new requirements for visual assets. High-quality, distinctive product imagery becomes even more important in this compressed visual format, requiring brands to consider how their products appear in these AI-curated collections rather than on traditional e-commerce pages.

5. First-party data strategy

OpenAI's stated intention to "explore ways for merchants to provide us their product feeds directly" signals the growing importance of first-party product data. Brands that develop capabilities to provide accurate, comprehensive product feeds directly to AI platforms may gain advantages in how their products are selected and displayed.

This non-advertising approach represents a significant departure from conventional digital marketing. While search engines and e-commerce platforms traditionally generate revenue through sponsored placements, ChatGPT's model prioritizes relevance and user experience with independently selected results. This could pressure established players to reconsider their reliance on advertising-driven product recommendations.

Timeline

• April 28, 2025: OpenAI announces ChatGPT's new product selection features and methodology
• April 28, 2025: Documentation released detailing the four key factors in product selection: structured metadata, third-party content, model responses, and safety standards
• April 28, 2025: Rollout begins of "visually rich carousels" and synthesized product information
• May 1, 2025: OpenAI reveals plans to develop direct merchant integration for product data feeds
• May 3, 2025: Current date, with product selection features continuing to roll out globally