All posts
Strategy7 min read

Share of voice in LLMs: a new metric for brand marketers

Share of voice is back — but the medium changed. Here is how to define, compute and act on share of voice across AI answer engines.

LE
LumenEntity Research
Visibility & AI search team

Share of voice (SOV) is the cleanest top-of-funnel metric for the LLM era. It compresses dozens of weaker signals into one comparable number.

Formal definition

On a fixed prompt panel P, over engines E, SOV_brand = sum(mentions of brand) / sum(mentions of brand + competitors).

Build the panel

Reuse the four-bucket prompt panel from your monitoring program. Lock it for at least a quarter to keep numbers comparable.

Weighting

  • Weight by prompt commercial value (branded < category < comparison).
  • Weight by engine usage if you have referral data.
  • Cap any single engine at 50% of the index to avoid distortions.

How to use it

Pair SOV with sentiment. A 40% SOV with 60% negative sentiment is worse than a 25% SOV with 90% positive sentiment. Plot both on a quadrant.

Frequently asked questions

Is SOV better than ranking?
It is a complement, not a replacement. Both belong in the weekly review.
How do I avoid panel bias?
Include competitor-flavored prompts. If you only track flattering ones, your SOV is inflated.
Can SOV predict revenue?
Branded-search SOV correlates with revenue. Category SOV is a leading indicator.
BrandMetricsGEO

Keep reading