TL;DR
Plenty of brands look like they are winning AI search — but they rank on a broad question the model is too blunt to resolve, for a market they do not actually own. We call it the coarseness premium. As models sharpen and agents query with the buyer’s exact spec, it gets collected. Ranked today, delisted tomorrow.
A brand looks like it is winning AI search. Ask an answer engine a broad category question and there it is — named, featured. Then ask the precise question its real buyer would ask, with the specific job, the specific surface, the specific constraint, and it is gone. Not a glitch. A coarseness premium.
It is one of the most common and least understood patterns in AI search, and it is quietly inflating a lot of brands’ confidence in a position they do not actually hold.
What a coarseness premium is
“Coarse” means blunt — low resolution. Today’s answer engines cannot yet tell fine market distinctions apart, and they rarely stop to ask the clarifying question a good salesperson would. So when a buyer asks something broad, the model reaches for everything adjacent and lumps it together. Brands that serve a neighboring market get swept into the answer. That undeserved inclusion — credit earned not by owning the market but by the model’s inability to separate it from the market next door — is the coarseness premium.
The cleanest analogy is a loosely graded test. Loose grading inflates the score. The score feels real until a stricter grader arrives.
We watched it across an entire market
We mapped an entire emerging category and compared who ranks on the broad category question against who is actually positioned in that specific market. A meaningful share of the brands surfacing on the category question are not in the category at all. They sit in adjacent markets and catch the broad spillover. On the specific question their buyer would really ask, they do not appear.
Those brands, if you asked them, would tell you AI search is going well. They see themselves featured. They have no way to see that the feature is borrowed.
Why it will not last
Two forces collect the premium, and both are accelerating.
The first is obvious: models get more precise with each release. As they learn to separate one specific use case from a neighboring one, and the buyer a product is built for from the buyer it merely touches, the adjacent brands stop qualifying for the specific answer.
The second is bigger. Buying is shifting to AI agents, and an agent does not guess. When a human types a broad query, the model fills in the blanks coarsely. When an agent shops on behalf of a buyer, it already carries the spec — the exact surface, the exact constraint, the exact use case — and it queries precisely, or it asks the clarifying question the human never did. Agent-mediated buying is spec-precise by nature. It is the thing that comes to collect the coarseness premium.
The question to ask about your own ranking
If your brand shows up in AI answers, the comfortable question is “am I featured?” The useful question is harder: am I featured for a reason that survives the next model — or only because the AI cannot tell yet?
If the answer is the latter, the ranking is a sugar high. You are ranked today and delisted tomorrow, and the drop will feel sudden even though it was always coming.
What to do about it
The work is not to chase the broad query harder. It is to know whether you actually own the specific market — and if you are riding spillover from an adjacent one, to close that gap while the premium still buys you time. The brands that win the next model are not the ones featured today. They are the ones who genuinely own the market when the answer engine finally learns to tell the difference.
That is a positioning problem before it is a content problem, and it stays invisible until someone measures the market at the resolution the next model will. Measuring it there is the discipline behind Genesis, the way Trendscoded starts every market: build your market — define the boundary precisely — and find out where you are actually positioned in it, who owns it and who is only adjacent. The coarseness premium is real, widespread, and carries an expiry date most brands cannot see — unless they are already looking at the market the way the next model will.
Frequently asked questions
What is a coarseness premium in AI search?
It is undeserved inclusion in an AI answer — credit a brand earns not by owning a market, but because today’s models are too blunt to separate it from the adjacent market next door. The brand ranks on a broad question for a market it does not actually hold.
Why will the coarseness premium disappear?
Two forces collect it, and both are accelerating: models get more precise with each release and stop lumping adjacent markets together, and buying is shifting to AI agents that query with the buyer’s exact spec instead of a vague human query.
How do AI agents change AI search rankings?
A human types broadly and the model fills the blanks coarsely; an agent already carries the spec — the exact surface, constraint, and use case — and queries precisely. Agent-mediated buying is spec-precise by nature, so adjacent brands stop qualifying for the specific answer.
How do I know if my AI ranking is real or borrowed?
Ask the harder question: am I featured for a reason that survives the next model, or only because the AI cannot tell adjacent markets apart yet? If it is the latter, the ranking is a sugar high — ranked today, delisted tomorrow.
What should brands do about it?
Do not chase the broad query harder. Determine whether you actually own the specific market; if you are riding spillover from an adjacent one, close that gap while the premium still buys you time. The brands that win the next model genuinely own the market when the engine learns to tell the difference.
