We measured 10 Brazilian sectors to find out who AI cites

Where these numbers come from

In August 2026 we measured ten Brazilian markets. We put the buying questions a person actually types when deciding where to buy in front of the engines, and classified 192 page-one positions by hand: brand site, marketplace, price comparator, press, forum.

In the same period we tested whether the sites of each sector's five largest brands return content to an automated reader. The result matters because an assistant that cannot read your site has nothing to cite you for beyond what other people wrote.

It is our own measurement, done by hand, with a date on it. Not a market survey copied from somebody else's report.

The main finding: there is no single rule

Market conversation about AI search tends to be generic. "Brands are being replaced by marketplaces", "AI prefers third-party content". The measurement shows this depends entirely on the sector, and the gap between the extremes is wide.

In pharmacies, 13 of 20 positions were pharmacy sites. The brand does hold its own ground. In footwear, exactly 1 position out of 20 belonged to a footwear brand; the rest were multi-brand retailers and list sites. Two retail sectors, same country, same month, opposite results.

Between those extremes, each market has its own dynamic:

• Automotive: 19 distinct domains across 20 positions. Nobody repeats, nobody has consolidated anything.
• Education: 7 of 20 positions are scholarship platforms and course comparators, intermediaries reselling the institution's own seat.
• Food retail: the most frequent domain in the sector was not a chain at all, it was a forum.
• Furniture: on the question that asks for brands by name, all five results were ranking articles, two of them published by retailers that resell the brands they listed.

The question that separates sectors most

One pattern showed up strongly: questions that ask for brands by name are almost never answered by brands.

In fashion, the theme that asks for Brazilian brands returned a blog, a magazine, an Instagram post, a YouTube video and a luxury marketplace. Not one brand site. In furniture, the same theme returned five ranking articles and no brands. In footwear, two of the measured themes asked for brands and were answered by multi-brand retailers and a review site.

It makes sense once you think about how those texts come to exist. A "best brands" list is content someone wrote on purpose to be found. A brand's own page talks about the brand, not about its category. Whoever wrote about the category won the category.

Technical readiness does not explain the result

The comfortable hypothesis would be that absent brands have bad sites. The measurement does not support it.

In footwear, all five of the largest brands passed both automated-reading tests. It is the best technical result we found in any sector. And the brands still held 1 of 20 positions.

In cosmetics, only one of the five leaders passed both tests, and the brands were absent too. So: a bad site hurts, but a good site alone puts nobody in the answer. These are two separate problems, and solving the first does not solve the second.

Social and forums: 9% of the total

Across the ten sectors, 17 of the 192 positions were social platforms or forums. A modest number in aggregate, but the distribution is a surprise: Reddit appeared in six of the ten markets, Instagram in three, YouTube in one.

That inverts the intuition of anyone investing in social for AI visibility. The conversation the assistant reads most often is not the one the brand publishes, it is the one customers write to each other.

What this means for your brand

The practical conclusion is uncomfortable for anyone looking for a recipe: there is no single one that works. The gap between pharmacy and footwear is too wide to extrapolate, and a plan built from a generic article will be right in one sector and badly wrong in another.

What works is starting from your own market. Finding out who occupies the space of the questions your customers ask, and how much of it is yours. Without that starting point, any content investment is a bet.

Which is exactly why every brand that works with us starts with a measurement of its own sector rather than an off-the-shelf plan.

The public breakdown of each market we studied is on the sector pages: cosmetics, pharmacies, fashion, footwear, food retail, automotive, education, furniture, clinics and software.

Limits of this measurement

Worth saying what it is not. It is a sample, not a census. The results are from August 2026 and will change, because page one changes. The reading was done for the Brazilian market, with results in Portuguese.

What it is: our own measurement, classified by hand, with a date on it. And it describes ten markets, not yours.

If you want to know what surfaces today when someone asks about your brand, the first check is free and takes under a minute.