How do I track AI share of voice?

Why it replaced keyword rank

There is no position one in an answer. An engine returns a paragraph naming a small handful of businesses, so the meaningful question is not where you rank but how often you are one of the names.

Share of voice captures that directly and is comparative by construction: it tells you not only how you are doing but who is taking the space you are not getting.

How to build the measurement

The method is simple; the discipline is what makes the number trustworthy.

• •Fix a question set of unbranded buyer questions covering your category and service area
• •Never include your brand name in the questions
• •Run the identical set across every engine you care about
• •For each answer, record every brand named, including competitors you did not expect
• •Compute your share as your named answers divided by total brand mentions
• •Repeat monthly with the same questions, and version the set if you ever change it

Separate named from cited

Being mentioned and being mentioned with a source pointing at you are different outcomes. Cited is stronger: it means the engine both recommended you and attributed the claim to you, which is what actually produces a visit.

Track them separately. A rising named count with a flat cited count usually means your brand is recognised but your site is not trusted as the source - a schema and corroboration problem, not a content one.

The mistakes that ruin the trend

Almost every broken share-of-voice dataset fails for the same handful of reasons.

• •Changing the question set between runs, so the trend measures the questions rather than your visibility
• •Testing only ChatGPT and generalising to every engine
• •Using signed-in sessions where personalisation flatters you
• •Sampling too few questions, so one competitor page swings the result
• •Recording only a score and discarding the answer text, leaving nothing to diagnose

Doing it at scale

By hand this is a spreadsheet and a recurring calendar block, and it works if you are consistent. It stops being practical at real volume - 100 questions across seven engines is 700 answers per run.

That is what MyFast AI automates: the same fixed question set, all seven engines, every verbatim answer stored, share of voice against the competitors you name, and progress tracking across every past report so the trend is comparable.