ChatGPT Shopping Results Now Favor Product Feeds Over Web Search

A dramatic shift in how ChatGPT selects shopping recommendations has left thousands of online retailers scrambling. Data from Profound, a company that runs customer prompts through ChatGPT daily and tracks the results, recorded one of the sharpest shifts in e-commerce visibility in recent memory — feed-integrated sources jumped from 8% to over 61% of product picks in a single day.

Across 1,757,723 tracked prompt runs in July 2026, feed-integrated product recommendations surged from 8.26% to 61.54% on July 10 alone. For store owners who had built their digital presence around traditional web search signals, the change arrived without warning and with little official explanation from OpenAI.

---

What the Data Actually Shows

Profound labels each product recommendation based on whether ChatGPT retrieved it through web search or through what it calls "feed-integrated retrieval." The company samples 50% of its daily prompt runs and counts each prompt once per day to build its tracking series.

The numbers tell a striking story. By August, feed retrieval had overtaken web search entirely across a broader sample of 97,725 prompt runs covering July 1 through August 24. As of September 3, Profound's tracking shows feed retrieval accounting for approximately 65% of the product recommendations it monitors.

Visibility Consequences Were Immediate — and Uneven

The impact on individual retailers varied considerably. Out of 687 customers in Profound's sample:

• 450 experienced at least a one-third reduction in their Shopping visibility when comparing July 7–9 with July 10–12
• 67 stores saw their visibility increase by that same magnitude during the same window

A statistical model Profound built from 517 customers who experienced significant visibility changes was able to explain 83% of the variation in outcomes. The model examined how much each store lost in web-search retrieval and how much it gained in feed retrieval — a level of explanatory power that points to a strong association between feed access and visibility outcomes within Profound's dataset.

Concentration of Visibility Among Fewer Merchants

The shift also concentrated visibility among a smaller group of merchants. The share of references going to the top 10 stores jumped from 22.5% to 41.8%, while the total number of unique merchants referenced fell from 13,524 to 10,607 — a drop of more than 20%.

This is a defining moment for e-commerce access: if your catalog isn't already connected through an approved feed, getting inside has become considerably harder. Retailers who have historically relied on organic discovery signals should note that effective e-commerce search engine optimisation now needs to account for structured feed data alongside traditional ranking factors.

Profound also estimates that roughly 35% of all feed retrieval in July was tied to Shopify, based on how closely the two daily data series moved together.

---

The GPT-5.6 Connection

OpenAI released GPT-5.6 on July 9, stating the rollout would occur over the following 24 hours. Profound's report attributes most of the July changes to this release, given the near-perfect timing alignment.

However, OpenAI has not publicly confirmed this connection. Its ChatGPT release notes include no shopping-related updates dated July 9 or July 10. The company's Help Center does state that ChatGPT selects products using structured data — such as price and description — from data providers and from stores themselves. It also states clearly that "product results are selected independently by ChatGPT and are not ads, nor influenced by any OpenAI partnerships."

Why This Matters for How Products Are Found

The mechanics of how AI models surface products are becoming increasingly distinct from how search engines rank web pages. Where traditional platforms rely heavily on signals like backlinks, on-page optimisation, and click behaviour, feed-integrated retrieval depends on the quality and completeness of structured catalog data. Understanding how to optimise individual product pages for search and AI visibility is becoming an essential capability for any retailer competing in this environment.

This divergence also has implications beyond ChatGPT. Shifts of this kind often signal broader changes in how AI-powered platforms evaluate and rank product data — a pattern worth monitoring across channels. For broader context on how AI is reshaping product discovery, Search Engine Land's coverage of AI in e-commerce provides ongoing industry analysis worth following.

---

Who Is Connected and Who Is Waiting

The Agentic Commerce Protocol

OpenAI's product feeds operate through the Agentic Commerce Protocol, its standard for sharing product data with ChatGPT. The protocol was extended to product discovery on March 24, 2026, with feeds available through providers including Salesforce and Stripe.

Shopify sellers already have their product catalogs integrated through Shopify Catalog with no additional setup required. Etsy catalogs are also connected. For everyone else, the path forward involves applying for direct feed access through OpenAI's merchant page — but applicants are currently placed on a waitlist.

OpenAI has not published data specifying how often ChatGPT relies on feed data versus open web retrieval, and that information was not available as of the publication date.

What's Coming Next

Looking ahead, OpenAI says a new self-serve platform is planned for later this year that would allow stores to connect their own feeds without going through the waitlist process. The company also says Shopping will expand to more regions over time.

Profound anticipates that the next competitive frontier will shift from simply having a connected feed to optimising the specific fields that feed contains — pricing accuracy, descriptions, and structured data quality. Additionally, OpenAI began rolling out GPT-6 Astra to a limited set of organisations on September 3, meaning the specific model answering prompts may vary when comparing results across different timeframes.

Practical Steps for Retailers

For retailers and marketers processing this shift, three considerations stand out:

1. Confirm whether your store's product catalog is already connected through an approved provider such as Shopify or Etsy — this may determine whether your products appear in ChatGPT Shopping results at all.
2. Apply for direct feed access through OpenAI's merchant page now — even with a waitlist, this positions you ahead of the self-serve platform launch expected later this year.
3. Audit your product feed data quality — pricing accuracy, descriptions, and structured fields may matter more to ChatGPT visibility than traditional SEO signals going forward.

Retailers who have been tracking performance through platforms like Google Shopping should be aware that Google Shopping performance insights can offer a useful benchmark for understanding how structured product data influences visibility across AI and search platforms alike. The principles of feed quality and catalog completeness are proving relevant across both environments.