We Measured AI Visibility for Two Industries. The Same Four Patterns Showed Up.

We ran the same AI-visibility audit across two industries that have almost nothing in common: dietary supplements and beauty. The four patterns we found in one showed up again in the other, at nearly the same shape and size. When a finding repeats across two unrelated categories, it stops being a quirk of one industry and starts looking like a rule of how AI search actually works.

This is first-party research, not a survey of opinions. We asked ChatGPT thousands of real buyer questions, recorded every brand it named and every source it cited, and counted. Two studies, run with the same generative engine optimization method, a few weeks apart. Below are the four patterns, the side-by-side numbers from both industries, and what to do about each one if you are the brand trying to get named.

How we measured it

Both studies used the same method: a fixed set of buyer questions per category, asked to ChatGPT while logged out from a US location, with every brand mention and every source link recorded and classified. Same 48 questions per category, both times. Logged out is the point, because that is the answer a first-time buyer sees, with no account history or personalization tilting the result.

Two things to keep straight before the numbers start. First, both studies are ChatGPT only. We are not claiming these exact figures hold on Perplexity, Gemini, or Copilot; those engines retrieve differently and would need their own runs. Second, the two industries were chosen because they are different on purpose. Supplements are a trust-and-safety, ingredient-driven category. Beauty is a look-and-feel, routine-driven one. Different buyers, different language, different shelves. If the same patterns hold in both, the pattern is about the machine, not the product.

Pattern 1: Most brands never get named

In an AI answer, most brands in a category are never named even once. There is no long tail to fall back on.

In supplements, 350 of the 500 brands we tracked were never named in a single answer. In beauty, 244 of 589 never came up once. That is the part most brands underestimate, and it is the core reason so many sites are invisible to AI. Classic search has a long tail, where page two and page fifty still exist and still pull the occasional click. An AI answer has no page two. It returns a handful of names and stops. If you are not in that handful, you are not lower down the list, you are absent from the answer entirely, and there is no scroll that rescues you.

The same fact is also the opening. Every category has shelves that no brand owns yet. The cheapest citation to win is the one nobody else is competing for.

Pattern 2: A short list repeats

AI answers default to a short, repeating set of brands, and the same names carry across categories that have little to do with each other.

The top three names took 49 percent of all mentions in supplements and 68 percent in beauty. A small set of brands is not just winning, it is absorbing the answer. And the same names keep reappearing in categories you would not expect them to. Thorne showed up across 20 of our 34 supplement categories. Clinique showed up across 26 of 35 beauty categories. These are not category specialists getting named for the one thing they make. They are default brands the model reaches for again and again.

The mechanism behind this is worth understanding, because it tells you how to compete. The model treats a name it has seen vouched for many times as a safe answer, so a brand that earns trust in one category gets handed the recommendation in adjacent ones it may not even lead. Trust compounds. That is bad news if you are starting from zero, and it is the reason the off-site work below matters so much.

Pattern 3: Brands vanish at the problem stage

AI names a brand far less often when the buyer describes a problem than when the buyer asks for the best product.

When the buyer asks “what is the best magnesium supplement,” ChatGPT named a brand 72 percent of the time in supplements and 93 percent of the time in beauty. When the same buyer instead described the problem, “I feel tired and wired in the afternoon” or “my skin is dry and flaky and nothing helps,” the brand naming rate fell to 11 percent and 29 percent. We named these the Symptom Void in supplements and the Bare-Face Gap in beauty, and they are the same hole in two industries.

Here is why it happens. A “best X” question is shaped like a list, so the model answers it with brand names. A described problem gets answered with advice: ingredients, routines, mechanisms, and often no brand at all. The buyer who is furthest from a decision, the one still describing a symptom in their own words, is exactly where almost no brand bothers to show up. That is ready demand going unclaimed, and it is wide open.

Pattern 4: Your reputation lives off-site

AI builds its brand recommendations from other people’s pages, not from yours.

A brand’s own website was the source behind the answer only 9 percent of the time in supplements and 2 percent of the time in beauty. Read those numbers again. The page you control, the one your team rewrites every quarter, is feeding fewer than one in ten supplement answers and one in fifty beauty answers. The rest comes from a small, repeatable set of third parties. In supplements, Reddit was the top source in 31 of 34 categories. In beauty, Allure led, with Reddit close behind.

This is the pattern that reorders a marketing plan. You can make your own site perfect and move almost nothing, because the model is reading the subreddit thread, the editorial roundup, and the comparison post, not your homepage. The work that earns an AI citation is mostly work that happens on pages you do not own.

What this means for your brand

Four patterns, four moves. None of them is exotic; the value is in knowing which one is costing you.

1. Do the off-site work. Your own site is the source behind fewer than 1 in 10 answers in supplements and 1 in 50 in beauty. Spend your effort getting named and reviewed on the pages the model actually reads: the relevant subreddit threads, the editorial roundups, the comparison posts. That is where the citation comes from, so that is where the work belongs.
2. Own the problem stage. The “best X” question is already crowded with default brands and hard to break into. The buyer who describes a symptom is barely served by any brand at all. Publish content that answers the problem in the buyer’s own words, and you are competing for a citation almost nobody else is contesting.
3. Claim an open category. With 70 percent of supplement brands and 41 percent of beauty brands never named, most categories still have room. Find the specific category questions where no clear default has formed yet, and become the obvious answer before the short list hardens around someone else.
4. Measure with a panel, not a spot check. One prompt, run once, tells you almost nothing, because the answer shifts from run to run and from engine to engine. You need a fixed set of questions, asked repeatedly, across the categories and the stages that matter, so you see the pattern instead of a single lucky or unlucky screenshot. That is the difference between checking and measuring citation readiness.

Where these numbers came from

These two studies were not a one-off research project. They were BlueJar’s own audit methodology, the same one we run for a single brand, pointed at two entire industries instead of one company. That is why the four patterns are not specific to supplements or beauty. They are how AI answers get assembled, in any consumer category where buyers ask an engine what to buy. They line up with what we see in the broader state of AI search in 2026.

Want to know which of the four patterns is costing your brand the most? Run your AI visibility audit at bluejar.ai across ChatGPT, Perplexity, Gemini, and Copilot. You will see where you are named, where you are cited as a source without being named, and where you are invisible, category by category.