This article was AI-generated as part of an experimental historical-content project. The date reflects the period being analyzed rather than the date the article was originally written.
Share of voice is one of the oldest numbers in communications. Count your coverage, count your competitors’ coverage, compare. Later it absorbed search: how many first-page results for a category term belong to you. The metric was never perfect, but boards could read it, because it counted things that could be counted.
Google I/O last week makes me wonder whether that number still measures the room where first impressions form.
The case for asking
The figures are Google’s own, but they are large. In his keynote, Sundar Pichai said AI Overviews have scaled to more than 1.5 billion users and are available in 200 countries and territories, and that the Gemini app has more than 400 million monthly active users. AI Mode began rolling out to everyone in the U.S. as its own tab. Outside Google, ChatGPT, Perplexity and Claude all search the web to answer questions, and Meta’s new standalone Meta AI app can search the web too.
None of these surfaces appears in a traditional share-of-voice report. A company can lead its category in coverage and still be absent, or described secondhand through a competitor’s framing, when someone asks an answer engine which firms to consider. In 2023, when generative search first went international, I argued that your first impression may not be a blue link. The scale numbers from last week make that harder to treat as a niche concern.
The case for caution
There are good reasons not to rush a new KPI into the board pack. Answers vary by phrasing, by session, by location and, increasingly, by personal context. A percentage for “mentions in AI answers” implies a precision the underlying data does not have. Vendors will offer exact-looking numbers, and some will be more exact than the answers deserve. Boards already see enough metrics that measure activity rather than outcome.
There is also the risk of chasing the wrong goal. Being mentioned is not the same as being described accurately. A company that appears in every answer because of an old controversy has a high share of voice and a problem.
A board-readable version
My answer is yes, boards should see something, but not a single number. A workable sketch has four parts, reviewed perhaps quarterly.
Presence. For a fixed set of questions a customer, investor or job candidate would ask about the category, does the company appear in the answers from the main engines?
Accuracy. When it appears, are the core facts right: leadership, what the company does, the status of any past issue?
Sources. Which sources do the answers cite: owned pages, major outlets, Wikipedia, forums, competitors? This is where the work happens, because sources can be improved and answers cannot be edited.
Comparison. How are the two or three main competitors described on the same questions?
Reported as a short table with directional changes and a few verbatim examples, this tells a board something real without pretending answers are stable. The fixed question set matters more than the tool that runs it, because consistency over time is what makes a trend meaningful.
Who owns it
This is the harder question. Search share of voice usually sat with marketing. Answer engines describe the company to investors, regulators and candidates as well as customers, which makes this a corporate affairs question as much as a marketing one. Whoever owns it needs authority to fix sources, not just report on them.
The old metric asked how loud a company was. The more useful question now is what the systems people ask actually say about it, and where they learned it.