AI Shopping Prioritizes Product Feeds Over Pages

AI-driven product recommendations, such as those presented in ChatGPT's carousels, are predominantly sourced from product feeds rather than individual product detail pages. A March 2026 study conducted by Tom Wells analyzed over 43,000 products featured in AI recommendations. The findings revealed that 83% of these products appeared within Google's top 40 organic Shopping results. In contrast, only 11% of products recommended by Bing matched Google's top results, and nearly all of these were also present on Google. This indicates that the products displayed by AI shopping assistants are not derived from the broader open web, product detail pages, or customer reviews, but rather from a specific file: the Google Merchant Center feed. Consequently, a brand's visibility in AI shopping is now directly dependent on the accuracy and quality of its product feed, diminishing the importance of product detail pages. The product feed has become the primary determinant of which items are presented to consumers. ChatGPT constructs its product carousels by utilizing "shopping query fan-outs," which are distinct from the search queries that generate textual answers. Wells' research demonstrated that one such query often involves retrieving a single page of Google Shopping results to populate an eight-product carousel, with 60% of strong matches originating from the top 10 Shopping results. The order of products within the AI-generated carousel directly mirrors their ranking in Google Shopping. Further analysis by Profound, which examined over 1 million ChatGPT shopping offers in June, yielded even more significant insights. Profound reported that approximately 99.9% of product citations directly pulled from merchant feeds were presented as the top product offer. The proportion of retrievals sourced from product feeds also saw a substantial increase, growing from 4.3% to around 20% of all ChatGPT shopping retrievals within a six-week period. This surge is attributed to the completeness of information provided by feeds. Profound's data indicated that feed-sourced offers consistently provided complete brand, product image, and merchant details 100% of the time. This contrasts sharply with page-scraped offers, which provided these details 0% of the time. Furthermore, feed-sourced offers were tagged with ChatGPT's "best price" designation 100% of the time, compared to only 21% for offers scraped from product pages.