For years visibility in search meant one thing: how far down the page you were appearing on the results page for a given keyword. That measurement still counts, but it doesn’t tell the whole story anymore. More people are getting their answers directly from AI chat tools, voice assistants and generative search summaries, and a business can rank well on a traditional results page and be entirely invisible inside those AI-generated answers. “AI share of voice” is the emerging term for how often, and how favorably, a brand appears when AI tools answer questions relevant to that business. It’s a new metric and most companies don’t have a good way to measure it yet. 

Why measuring this is harder than for traditional rankings 

Traditional rank tracking works because search results are stable enough and easy to query at scale. Responses generated by AI are not uniform in the same manner; the same question can produce different answers depending on phrasing, timing and what model is used. There is no one “position one” to follow. Rather, the useful signal is more like: how often is the brand mentioned at all, across a representative set of realistic questions, and how is it described when it is? 

A practical way to keep tabs on it 

1. Create a list of 20-30 realistic questions that a potential customer might ask an AI assistant about the category of the business – not just the business name. 

1. Run those questions periodically across two or three major AI tools and note if the brand appears, how it’s described, and which competitors appear instead. 

1. Pay attention to which sources the AI tool cites. When it shows its reasoning or references, this can often give insight into which of the brand’s own pages (if any) are being pulled from. 

1. Watch these over months, not weeks. Generative answers may change as models are updated. Short-term snapshots may be noisy. 

1. Compared to a small set of direct competitors for relative position, not just absolute presence. 

How to use the results 

For a brand that appears infrequently, the fix is likely the same as what helps traditional AI citations: clearer, more specific, more original content that answers the exact questions being asked, published in a format AI tools can parse cleanly. Where a brand is present but named incorrectly, it’s a clue worth checking out separately, an out-of-date reference somewhere online may be the source. 

This kind of tracking is still a new frontier, and no tool measures it perfectly yet. But companies that start building even a rough, manual version of this dashboard now will have a real edge on competitors who only notice the shift once their traditional traffic is already beginning to fall away.