AI share of voice scores mislead brands, Taylor warns

A polished AI share-of-voice percentage can look like boardroom proof, but Dan Taylor says it often disguises weak math. His warning targets the new AI visibility market, where tools turn a tiny slice of prompts into a neat-looking market-share number and present it as hard evidence.

That is the flaw Taylor keeps returning to. Classic share of voice worked because teams could lock down a fixed keyword set and know the denominator. AI search does not behave that way. The prompt universe is effectively endless, answers are dynamic and personalized, and different platforms retrieve information in different ways. Once a tool extrapolates from a small sample of prompts, a crisp percentage can create a false sense of precision that executives and finance teams will happily overread.

Taylor’s answer is to stop worshipping one dashboard number and start measuring three things that map to real outcomes. First is prompt-cluster coverage, whether a brand shows up in the questions that actually matter, which is the clearest read on discoverability. Second is recommendation rate, not just raw mention volume, because being cited in passing is not the same as being pulled forward as a buying option. Third is movement over time across the platforms that shape decisions, which tells you whether visibility is improving in a way that can plausibly affect business impact.

The broader research base points in the same direction. Princeton University researchers introduced Generative Engine Optimization, along with GEO-bench, to evaluate and improve visibility in generative engine responses. A later GEO paper found a systematic bias toward earned media over brand-owned and social content, and showed that AI search services vary sharply in domain diversity, freshness, cross-language stability and sensitivity to phrasing. That means a single global score flattens the very platform differences that matter most.

Ahrefs has put a hard number on that break from old-school SEO logic. In its updated analysis of 863,000 keywords and 4 million AI Overview URLs, only 38% of pages cited in Google AI Overviews also ranked in the top 10 for the same query, down from 76% in an earlier study. The message for brands is blunt: ranking well is no longer the same thing as being surfaced, cited or recommended, and a pretty percentage will not survive scrutiny if no one can explain the denominator.