Ten US categories. One hundred ninety-nine answers. Who gets named.

Ten buyer questions in each of 10 US categories, sent once to ChatGPT and once to Google AI Mode, from a VPN exit in Washington DC, so the market seen matched the market named in the question. 10 categories times ten questions times two engines is 200 answers; one, in supplements, was never captured, so this report counts 199. A mention is the brand's name appearing as a whole word in the answer text; every brand named was counted, not a pre-chosen set. Cited URLs were recorded separately from named brands, so cited but not named could be reported on its own.

This is the same method run once before, on a different pair of engines, in a different market. That earlier study is linked at the end of this page; the two are never compared on the same line, because it ran through search-grounded APIs and this one ran through the consumer apps.

Lead brand, how often it was named, the top two brands' combined share, and which brand each engine leads with, category by category.

All 10 categories returned a brand in every single answer; none came back empty. Gusto led payroll and HR software, named in 20 of 20. QuickBooks, Asana, HubSpot and Talkspace each led their own category in every answer too. The rest still led by a wide margin without quite reaching every answer: Bluevine was named in 17 of 20 business banking answers, for one.

An empty shortlist, the finding that led the India study, is not what the US data shows. Every category measured here already has an incumbent an engine reaches for first.

These ten categories are sold nationwide or online, so one shortlist can hold across the whole country. Local businesses work differently. For a car dealership, a plumber or a dentist, the answer depends on where the buyer is searching from, often down to the neighborhood, so no single company can own the category everywhere. That leaves room for a local business to be recommended in its own area. This study did not measure local categories.

In 6 of the 10 categories, the top two brands together were named in every answer. The lowest share was in supplements, where Thorne and Nature Made were named in 68% of answers between them.

A shortlist this concentrated is also the opening: the third and fourth brand in a category are splitting what is left over, not competing for the top spot.

9 of 10 categories had a top brand shared between the two engines, some only because of a tie. In accounting software for small business, ChatGPT tied QuickBooks and Wave and Xero at the top, while Google AI Mode led alone with QuickBooks. The two sets shared a brand, so the category counts as agreement, not as one tie broken in either direction. The one exception was pet food: ChatGPT's top brand was Purina, named in 7 of its 10 ChatGPT answers; Google AI Mode's was Hill's, named in 8 of its 10 answers. The two sets shared nothing.

A shared brand inside a tie is not the same claim as the same brand leading outright: where an engine ties several names at the top, any one of them could be what a buyer actually sees first. Only pet food showed two genuinely separate leaders, and the margin there was 1 answer either way.

Airwallex's own domain was cited in 11 of the business banking answers, more than any other source. The brand itself was named in only 5 of 20 answers. The engines read the page and still recommended someone else more often than not.

Most of the domains cited often in this sweep are not brands at all. Reddit and Forbes are cited across more than half the categories measured, and none of them is a product an answer could name.

• 199 of 200 planned answers. Supplements, ChatGPT, question 3, "Which supplement brands in the United States are third-party tested?", was never captured.
• One run per question. Answers vary between runs; these are a reading, not a constant.
• One US location: a Washington DC VPN exit. A buyer in another city may see a different answer.
• The market was named in every question, and Google's gl parameter was set to match.
• Ten questions per category, written by us. A different question set would move individual numbers.
• Consumer apps: ChatGPT temporary chat and Google AI Mode, captured 8 September 2026.

This is the same method run once before, in a different market on a different pair of engines. an earlier study in one market.