Go looking for proper numbers on how brands show up in AI answers — in a category we know well, supplements — and the picture that comes back is blunter than expected.
Across roughly 3,800 supplement prompts tested between January and April this year — run against ChatGPT, Claude, Perplexity, Gemini and Google’s AI Overviews — four brands took an estimated 47% or more of observed citations between them: Thorne, Seed, AG1 and Momentous.
A separate benchmark, narrower in scope, found the same shape inside multivitamins alone: a handful of brands taking the clear majority of mentions, and a substantial minority of the brands tested not appearing at all. Not lower down the list. Not occasionally. Zero.
Sit with that. A meaningful slice of a category worth tens of billions is completely absent from the moment a buyer asks the question that matters most: “what should I actually buy?”
Search was never this brutal
The whole logic of SEO was the long tail — a long tail of brands ranking for a long tail of terms. You could be a small player, pick your niche keywords, sit on page two for the broad ones, and still get found. The page was generous: ten blue links, then more below, then more after that. Being twelfth was survivable.
AI answers don’t have a page two. When someone asks “what’s a good magnesium for sleep?”, the assistant doesn’t return a ranked list of forty options to scroll. It returns a shortlist — three names, maybe five. You are on it, or you don’t exist in that conversation. There’s the answer, and there’s everything the answer left out.
That changes the shape of the competition completely. Search rewarded effort spread across a thousand terms. AI discovery rewards being the obvious answer to the question your buyer actually asks. It’s winner-take-most, and it resolves much faster than search’s slow rank decay ever did.
What decides the shortlist
Here’s the part brand teams find hardest to accept: what decides the shortlist isn’t who shouts loudest or spends most. Some of the invisible brands in the data outspend the visible ones.
What decides it is which brands are legible to the model — consistent claims across their own site and the places the AI trusts, comparison-ready against named alternatives, with public, checkable evidence the model can lean on without taking a risk. The assistant is trying not to be wrong. It reaches for the brands it can stand behind.
Most of that work is unglamorous. It’s not a campaign. It’s making sure that when an assistant checks three sources about your product, it finds the same dose, the same testing claim, the same story every time — instead of three slightly different versions of you it can’t reconcile, so it quietly picks someone else. That’s the coherence problem in one sentence.
The silent shift
The real shift underneath the numbers: discovery is moving from the search bar to the assistant, without a clear before-and-after moment to mark it. There’s no impressions column for an answer you weren’t included in. No alert fires. You don’t lose the shortlist seat loudly — you never had it, and you find out when your traffic does.
The surprising thing is how few brand teams have looked. It takes thirty seconds: open the assistant, ask the question your best customer would ask, read the answer as if you were them. It is reliably the most uncomfortable thirty seconds of the week — and the most useful. You can’t fix a gap you’ve never seen.
The brands that notice early get the seat. The rest will spend next year wondering where the demand went.
Benchmark data drawn from the 2026 Supplements AI Visibility Index (5WPR), with supporting multivitamin-category benchmarking from Avenue Z.
Find out if you’re on your category’s shortlist: get your free GEO Score.