AI search analysts at Peec AI published a review of ChatGPT's product recommendations in two markets. According to their data, in the US OpenAI ranks products using its own Labrador engine. In Europe, the analysis suggests, ChatGPT relies on Google Shopping results.
The sample, as described, is large: about 150,000 rows of debug data. We did not verify the methodology or the raw data, so all conclusions below belong to Peec AI. OpenAI has not publicly confirmed these differences, at least not in any statement we have seen.
What distinguishes the US market, according to Peec AI: a store without verified seller status drops out of the carousel. The product description is considered roughly within the first 40–50 words (50 tokens). Different models and colors may occupy separate positions in the feed.
In Europe, as the analysis indicates, OpenAI does not have its own seller scoring. The entry threshold for new stores is lower, and positions in ChatGPT depend on how well a standard feed is configured for Google Shopping.
Peec AI also writes about fan-out queries, which the model appends to the user's request on its own, such as "best", "review", and "comparison". If these terms are absent from the feed or the landing page, a listing risks not making it into the carousel. For products rated below three stars, according to the same data, visibility drops sharply.
What remains unclear: how long this scheme has been in place, whether it changes as the feature rolls out, and how stable the results are when prompts are varied. A single analytical source and unpublished OpenAI details are a reason to test your own feeds with your own sample queries, rather than rebuilding them entirely based on this review.
