How often do AI answers name no businesses at all?

Of the 926,792 answer units examined in this study, spanning 193 query groups over the window 2026-04-27 to 2026-09-18, 2.3% named no businesses at all. This class, names none, is by far the smallest of the four classes measured, well below names 1 to 5 at 6.6%, names 6 to 9 at 51.8%, and names 10 or more at 39.4%. In plain terms, when an AI engine is asked a question in this study's query set, it names at least one business in the overwhelming majority of its answers.

But the rarity of a zero-business answer is not uniform across engines. Gemini produced a names none answer 5.7% of the time, out of 298,038 answers. ChatGPT produced one only 0.4% of the time, out of 329,786 answers. Claude sat between the two, at 0.9% of its 298,968 answers. Gemini's rate is well over an order of magnitude higher than ChatGPT's, even though both rates are small in absolute terms.

This matters for a specific practical reason: a names none answer is the one outcome in this classification where a business has zero chance of being mentioned, because nothing was mentioned. If your visibility strategy depends on being one of the businesses an AI engine names, the names none rate for a given engine sets a ceiling on how often that engine will ever produce a name-bearing answer to begin with. A higher names none rate, like Gemini's, means a larger fraction of queries end without any business being named by that engine, regardless of which business it might have been.

Why would an engine decline to name any business at all? Several explanations are plausible and this study cannot distinguish between them from the classification data alone. One is caution: an engine may be tuned to avoid naming specific businesses for certain query phrasings, particularly ones that read as seeking a recommendation rather than information. Another is simply that some queries in the group set do not call for naming a business at all, and an engine that names none there is answering correctly, not conservatively. A third is variation in how each engine interprets an ambiguous prompt, naming a general category or type of business rather than a specific one, which this study's classification would also record as names none since it counts distinct named businesses. Distinguishing caution from correct non-applicability would require inspecting the actual query text and answer content behind each names none classification, which is beyond what this counting exercise does.

The spread across query groups reinforces that query type, not just engine, drives this outcome. Across the 188 groups counted, the names none share ranges from 0.2% to 6.9%, with a median of 2%. Even the highest group-level rate is a small fraction of that group's answers, which tells you that no single query group is dominated by zero-business answers. The rarity of names none is a stable feature of this dataset, not an artifact of a few unusual groups pulling the average down.