What Share of Companies Never Appear in AI Search Results at All?

Before asking how well a company ranks in AI search, it's worth asking a more basic question: does it show up at all? This study tested a cohort of 58,803 companies by issuing 8,188 distinct prompts to three AI assistants, ChatGPT, Gemini, and Claude, and collecting 536,301 usable answers over a run window from September 4 to September 25, 2026. Every answer was checked for whether the company in question was named anywhere in it.

The headline number here is small but worth stating precisely: 0 companies in the cohort, 0% of the total, never appeared on any engine for any prompt. Framed the other way, 56,194 companies, 95.6% of the cohort, were named by at least one engine at least once.

That is a strikingly high floor. It suggests that for this cohort and this set of 8,188 prompts, near-total invisibility is rare. If a company has any kind of established presence that these engines' underlying training and retrieval processes can draw on, it is very likely to surface somewhere, for some question, on some engine.

This is worth sitting with because it cuts against a common fear, that AI search is a winner-take-all environment where most companies are invisible and a small number dominate every answer. That is not what this study found. What it found instead is closer to the opposite problem: almost everyone shows up somewhere, which means "we were mentioned by an AI assistant" is a low bar, not a differentiator. The harder, more useful questions are about how often a company appears, on which engines, in what position, and for which categories of prompt, not whether it appears even once.

Look at the category-level breakdown to see why the binary visible-or-not framing undersells the real story. Every category in this study, from large ones like E-commerce Solutions (3,600 companies) and Mobile App Development (5,139 companies) down to small ones like Tourism Digital Marketing (8 companies) and Fitness Digital Marketing (7 companies), shows a visible rate of 100% or its equivalent, essentially full coverage at the category level. That uniformity means category alone doesn't explain who gets left out. Something else, likely company-level factors like how much public information exists about a business, or how distinctively its name maps to the kinds of buying questions asked, is doing the work.

It also means the near-zero non-visibility figure should not be read as evidence that AI search is fair or evenly distributed. A company can clear the bar of "named at least once" while still being named rarely, named only by the engine with the least usage share, or named only far down in a list where a reader skimming an answer would never scroll to see it. This study's position data shows mentions average out to a position of 4.94 across 197,893 ranked mentions, with some spread by engine (5.08 on ChatGPT, 4.97 on Claude, 4.72 on Gemini). Being named once, in last place, on the engine fewest people use, counts the same in the never-appears tally as being named first, everywhere, constantly.

What would make this finding fragile is the prompt set itself. The 8,188 prompts used here were built to resemble realistic buying questions across categories in the cohort; a different prompt set, especially one weighted toward very narrow or hyper-local questions, could plausibly push the never-appears count up. The honest reading of 0% is: among companies that are plausible answers to at least one of these 8,188 questions, almost none are completely invisible. It does not claim that every company anyone might search for would clear the same bar.