The landscape of e-commerce discovery within generative AI is undergoing a fundamental transformation, as recent data from Profound reveals a seismic shift in how ChatGPT surfaces products to consumers. For years, search engines relied on the open web as the primary source for product discovery. However, analysis of over 1.7 million tracked prompt runs indicates that ChatGPT has pivoted aggressively toward a "feed-integrated" retrieval system. This structural change, which gained significant momentum on July 10, marks a departure from traditional web-crawling methods, favoring direct, structured data feeds from merchants. As OpenAI continues to refine its Agentic Commerce Protocol, retailers find themselves at a crossroads where the technical integration of their product catalogs may be the single most important factor in maintaining visibility within the world’s most popular AI interface.
The July 10 Inflection Point
The transition toward feed-integrated recommendations was not gradual; it was immediate and statistically profound. Prior to July 10, feed-sourced recommendations accounted for a mere 8.26% of tracked product recommendations. By the end of the day on July 10, that figure had surged to 61.54%. This shift aligns precisely with the rollout of GPT-5.6, an update OpenAI initiated on July 9 with a 24-hour deployment window.
While OpenAI has remained silent regarding specific shopping-related updates in its public release notes for that period, the correlation in the data suggests a deliberate architectural change in how the model parses shopping intent. Profound’s longitudinal analysis—which sampled 10% of prompt runs between July 1 and August 24—confirms that this was not a temporary glitch. By August, feed-based retrieval had officially overtaken web-search retrieval, and by September 3, feed-sourced results accounted for approximately 65% of all product recommendations tracked by the firm.
Impact on Merchant Visibility
The consequences for digital retailers were immediate and, for many, disruptive. Among a sample group of 687 merchants, 450 experienced a minimum 33% reduction in shopping visibility in the three-day window following the July 10 update, while a smaller cohort of 67 merchants saw an equivalent increase.
The volatility suggests that the algorithm’s new preference for feed-integrated data favors stores that have successfully connected their catalogs directly to OpenAI’s systems. Furthermore, the retrieval mechanism appears to have become more concentrated. During this period, the diversity of the merchant ecosystem within ChatGPT’s results shrank significantly. The number of unique merchants referenced in prompt responses dropped by over 20%, from 13,524 to 10,607. Simultaneously, the top 10 most-referenced stores saw their share of total recommendations jump from 22.5% to 41.8%. This concentration indicates that feed retrieval draws from a narrower, more curated, and technically integrated pool of merchants compared to the broader, more democratic results provided by traditional web search.
Understanding the Agentic Commerce Protocol
To understand why this shift occurred, one must look at the technical infrastructure OpenAI is building: the Agentic Commerce Protocol. This standard enables ChatGPT to ingest structured product data—prices, descriptions, availability, and attributes—directly from providers.
Currently, the ecosystem is tiered. Shopify and Etsy, having established direct integrations with OpenAI, appear to be the primary beneficiaries of this shift. Profound’s data suggests that approximately 35% of all feed-based retrieval in July was tied to Shopify merchants, reflecting the platform’s deep integration with OpenAI’s catalog services. For other retailers, the path to visibility is more arduous. While OpenAI allows retailers to request direct feed access, many applicants remain on a waitlist.
OpenAI’s official stance remains that product results are selected independently and are not influenced by paid partnerships or advertisements. According to the company’s Help Center, ChatGPT utilizes structured data from providers and individual stores to make its selections. However, the lack of transparency regarding how often the model chooses to prioritize a direct feed over an open-web search result leaves many businesses in a state of uncertainty. As it stands, the "black box" nature of the model’s retrieval logic makes it difficult for merchants to optimize their performance, other than by ensuring their technical compliance with OpenAI’s data standards.
The Evolution of Search and Discovery
The shift from open-web search to feed-based retrieval represents a broader trend in AI development: the move toward "Agentic Commerce." In this model, the AI does not simply act as a search engine that directs users to a website; it acts as a concierge that processes structured data to make decisions on behalf of the consumer.
This evolution brings both risks and rewards. For consumers, the move to feeds may result in more accurate pricing, real-time inventory updates, and fewer broken links. For merchants, however, it changes the rules of the game. Search Engine Optimization (SEO) was historically about content, authority, and backlinks. In the era of AI-driven shopping, "Feed Optimization" may become the new standard.
The data suggests that the next phase of competition among retailers will not just be about having a feed, but about the quality and granularity of the fields within that feed. As ChatGPT becomes more sophisticated—exemplified by the rollout of GPT-6 Astra to select organizations—the model’s ability to parse complex product attributes will likely increase. Retailers who provide rich, structured, and accurate data will be better positioned to satisfy the model’s "intent-matching" requirements.
Implications for the Retail Industry
For digital marketers and e-commerce managers, the findings from the Profound report serve as a wake-up call. Relying on organic traffic through search engines is no longer a sufficient strategy for capturing AI-driven intent. The data suggests that if a merchant’s product data is not integrated via an approved channel—such as Shopify, Stripe, Salesforce, or direct feed submission—they are essentially invisible to a significant portion of ChatGPT’s shopping traffic.
Furthermore, the volatility observed in July underscores the dangers of platform dependency. When an AI model changes its retrieval logic, the impact on business traffic can be sudden and severe. With OpenAI planning to launch a self-serve platform later this year to allow more stores to connect their feeds independently, the barrier to entry may eventually lower, but the intensity of competition within those feeds will likely increase.
Looking Ahead: The Future of AI Commerce
As we move into the final quarter of the year, several factors will influence the future of shopping on ChatGPT. First, the regional expansion of the shopping feature will increase the total addressable market for merchants, likely bringing more global retailers into the fold. Second, the anticipated self-serve portal will democratize access, potentially reversing the current trend of merchant concentration by allowing smaller players to feed their data directly into the system.
However, the core challenge remains: transparency. As long as OpenAI keeps the weighting of its retrieval algorithms private, merchants will continue to operate on observation rather than instruction. The industry will need to rely on third-party observational data—like the reports provided by Profound—to gain visibility into these shifts.
The transition from web-search to feed-integrated retrieval is not merely a technical update; it is a fundamental shift in the economics of the internet. As ChatGPT continues to evolve from a chatbot into a primary commercial interface, the ability to deliver structured, high-fidelity product data will become the single most critical asset for any brand operating in the digital space. Retailers must now view their product catalogs not just as static web content, but as dynamic data streams that feed the intelligence of the next generation of commerce.
