ChatGPT Searches for Specific Brand Names in 52% of Shopping Questions

Someone asks ChatGPT for the best travel credit card for earning points. Before any results come back, ChatGPT's hidden search already reads "best travel credit cards points 2026 Chase Sapphire Preferred Capital One Venture X Amex Gold". It has a shortlist in mind, and it searches to check it.

That changes what it takes to appear in an AI shopping answer. In classic search, a brand competes for a ranking on the words people type. In ChatGPT, the brands in the hidden search are chosen before any page is read, so a brand that is not on that first list may never be searched for at all, while one that is gets its reviews and its own pages checked.

In 52% of ChatGPT's answers to "best product" questions, at least one hidden search names a brand. The searches often list several rivals together, and the longest chains end on the brands' own websites, checking pricing and fees directly.

On September 23, 2026 we asked ChatGPT, with web search switched on and from a US location, a set of "best product" questions, from credit cards and accounting software to robot vacuums and strollers. For 138 of the answers, ChatGPT's data included the web searches it ran in the background to write them.

A hidden search is a query ChatGPT sends to the web on its own while preparing an answer; the user sees the answer, not the searches. We checked each hidden search for a brand name from that product category and counted the answers in which at least one search named a brand. We also read the search chains to see what ChatGPT did with the brands it named.

That is 72 of the 138 answers. ChatGPT runs 1.6 hidden searches per shopping answer on average, and in half of the answers the brands are in the search from the start. The brand searches look like verification: ChatGPT names the products it expects to recommend, then looks for reviews and roundups that cover them.

The brands are in the search before any result comes back, so the shortlist starts from what ChatGPT already knows about the category. The search then tests that list against current sources. The year is usually there too: 66.5% of ChatGPT's hidden searches for "best product" questions add "2026".

The longest search chains go from a roundup to each brand's own website. Across all 602 hidden searches in the wider set, 32 used a site: search, the command that limits results to one website. For "best small company accounting software", ChatGPT ran 11 hidden searches: a roundup naming QuickBooks, Xero, FreshBooks, Wave and Zoho Books, then searches inside each vendor's pricing pages.

Credit cards work the same way. For "best credit card for reward travel", ChatGPT searched a roundup, then chase.com and capitalone.com for rewards and annual fees, including "site:chase.com Sapphire Preferred rewards annual fee transfer partners 2026". The roundup decides who is on the list; the brand's own site supplies the facts ChatGPT quotes.

Brand awareness inside the model now matters directly, because the brand names in these searches come before any results are read. If ChatGPT names your competitors in its first search and not you, you start behind. The lever is coverage in the reviews, roundups and comparisons that shape how the model sees your category, so that your name is on its list before it searches.

Then keep your own pricing and product pages ready to be checked. When ChatGPT searches inside your website, it wants current prices, plans, fees and specifications it can quote, in plain text on public pages. Working with an ai seo company shows which brands ChatGPT searches for in your category and whether yours is one of them.

Cite this study