Published
The 73% and the 11%
Research into how AI systems answer commercial questions produced a stratification that is difficult to argue with. Global household names appear in around 73% of relevant AI answers. Established mid-market brands appear in 44%. Niche brands appear in 11%.
The gap is not explained by page quality. Smaller companies frequently publish more specialised and more accurate material about their own field than the large brands that outrank them in AI answers.
It is explained by corroboration. Language models assemble answers from patterns across an enormous body of text. A company described in thousands of independent places is a strong, stable pattern. A company described in forty is closer to noise, and a system optimising for confident answers will reach for the pattern it trusts.
Analysis of winning and losing sites through 2026 found the same effect from the other direction: sites with strong brands won roughly twice as often as everyone else.
How an old separation broke
For twenty years, brand and search sat in different budgets and frequently different agencies. Brand governed how a company was perceived. Search governed how it was found. The two were understood to interact loosely and were measured entirely apart.
That separation no longer describes the mechanism. If a system's willingness to name a company depends on how widely and consistently that company is described elsewhere, then reputation has become an input to retrieval — not an influence on it, an input to it.
Google's guidance reinforces the point from the opposite side. It warns that pursuing manufactured mentions is unproductive because the systems weigh quality over volume. The corroboration has to be genuine, which rules out the tactics that would otherwise be used to fake it.
Why this is not a counsel of despair
Read carelessly, the 11% figure says small companies cannot compete. Read alongside the rest of the research, it says something narrower and more useful.
AI visibility behaves as a topic-level game rather than a single brand score. A company is not visible or invisible in general; it is visible or invisible on particular questions. The household names dominate the broad category terms because that is where their corroboration is thickest.
On a specific, narrow question — a niche process, a regional market, an unusual constraint — the corroboration advantage largely disappears, because the large brand has never written seriously about it either.
What a smaller company should actually do
The realistic objective is topic ownership, and it is a different exercise from category marketing.
Identify the two or three questions where the company has knowledge its larger competitors do not: proprietary data, unusual case experience, a market they do not serve well. Publish the evidence in detail, including the parts that complicate the story, because that is the material other writers cite.
Then pursue corroboration deliberately. Trade press, industry associations, podcasts, conference talks, community answers, partner content. The aim is for the company's name to appear near that topic in places it does not control — which is where 84% of AI citations originate.
Being the name that comes up for one specific question is achievable within a year. Being the name that comes up for an entire industry is not, and the budget spent attempting it is the budget that would have won the first.
The measurement that follows
Share of voice needs a second definition. Alongside the traditional media measure sits a simpler test: when a buyer asks an AI system the question that leads to a purchase, is the company named, and is the description accurate?
That has to be asked manually, at intervals, across several systems, because citation patterns diverge sharply — only 11% of domains cited by ChatGPT overlap with Perplexity's. It is slower than a rank check and closer to what is actually being bought.
The second measure is share of search, which tracks whether the underlying fame is growing. The two together answer the question a brand budget has always struggled with: is anyone hearing about us, and does the machine believe it?
Common questions
Why does my company never appear in AI answers?
Most likely a corroboration problem rather than a content problem. Niche brands appear in around 11% of relevant AI answers against 73% for household names, because AI systems favour companies described consistently across many independent sources.
Can a small business compete for AI visibility?
On specific topics, yes. AI visibility works topic by topic rather than as a single score, and large brands hold no corroboration advantage on narrow questions they have never written about.
Does brand building affect SEO now?
Directly. Brand recognition is an input to whether AI systems name a company, and strong-brand sites won roughly twice as often as others in 2026 analyses of search winners and losers.