Skip to content

Module 6 min

AI Visibility & SoV

Share of Voice = how often you appear across a defined prompt set versus competitors. It is the headline GEO metric.

28%rank #2
  • You28%
  • Rival A30%
  • Rival B22%
  • Rival C13%
  • Everyone else7%

SoV is relative: gaining share means taking it from someone. Rank is what leadership remembers.

11% ChatGPT/Perplexity citation overlap; track engines separately

Why it mattersSoV is relative: you gain only by taking share, and rank is what leadership remembers.

01

Definition & Foundation

What it is, in plain words

Share of Voice (SoV) is the headline GEO metric: how often you appear across a defined set of buyer prompts versus your competitors, expressed as a share. It answers the question leadership actually asks ("when our customers ask the AI, how often is it us?") and it converts a fog of individual citations into one number you can trend, benchmark, and defend a budget with.

Two properties make SoV different from a ranking. First, it is relative: it's a share of a finite answer, so you gain only by taking it from a rival; a rising SoV is always someone else's falling one. Second, it must be computed per engine, because the citation landscapes barely overlap; only about 11% of cited domains overlap between ChatGPT and Perplexity. A blended "AI visibility" number averages away the truth; you win Perplexity and lose ChatGPT with the same score. The craft is a representative prompt set, run on a schedule, scored per engine, benchmarked against named rivals.

02

Four Ideas Behind a Real SoV Number

What separates a defensible metric from a vanity one

the denominator

The prompt set

A representative, fixed list of the buyer prompts that matter, the universe you measure share of. Drawn from prompt research, it must stay stable enough to trend and broad enough to be honest, not cherry-picked to flatter you.

how you score

Prominence, not just presence

Being cited first or quoted directly is worth more than a passing link at the bottom. A good SoV weights how you appear (cited, mentioned, quoted, or absent), not just a binary "were we there," so the number reflects real influence.

never blend

Per-engine scoring

With ~11% cross-engine overlap, one blended figure hides which engine you're losing. Compute and report SoV separately for each engine you care about; the divergence between them is often the most actionable thing you learn.

share is relative

Competitive benchmarking

SoV is meaningless in isolation; it's your share versus named rivals. Track the same prompt set for your top competitors so a move in your number always has a "from whom / to whom" attached. That framing is what leadership funds.

03

Myths vs Reality

Common misreadings, corrected

Myth"One AI-visibility score across all engines tells us how we're doing."
RealityA blended score is an average that hides the split: with only ~11% citation overlap, you can be dominant on Perplexity and invisible on ChatGPT at the same "score." Per-engine SoV is the only honest version; the blend is a vanity number that obscures your biggest gap.
Myth"Our SoV went up, so we're winning."
RealitySoV is relative and zero-sum within a prompt set; it can rise because a rival stumbled, or fall while you improve because a competitor surged harder. Always read it against named benchmarks and the underlying citation changes, not as an absolute score that only goes up when you do well.
04

Putting It to Work

A repeatable, per-engine, benchmarked score

The work is turning "are we visible in AI?" into a number you re-produce the same way every cycle: same prompts, same engines, same scoring, with rivals in frame.

The Share-of-Voice playbook

1

Fix a representative prompt set

From prompt research, lock a stable list of the buyer prompts that matter across intents. Keep it fixed enough to trend; change it deliberately and note when you do, so a score jump is never just a changed denominator.

2

Run it across each engine on a schedule

Execute the full set on ChatGPT, Perplexity, AI Overviews, and any category-specific engine, on a regular cadence. Consistency of method matters more than frequency: same prompts, same way, every cycle.

3

Score presence and prominence

For each prompt/engine, record whether you're cited, mentioned, quoted, or absent, and how prominently. Aggregate into a per-engine share. Capture the source URL too; it feeds citation tracking.

4

Benchmark against named rivals

Run the identical set for your top 3–5 competitors and report share side by side. Every movement in your number should carry a "share taken from / lost to"; that's what makes SoV a strategic metric, not a vanity one.

5

Report it as the GEO scoreboard

Put per-engine SoV, with competitor benchmarks and trend lines, into the citation column of your dual scoreboard. This single artifact is what turns "is GEO working?" from an argument into a chart.

05

Verification Checks

How to know it's really done

0/4 verified

Your SoV is a real metric when it's reproducible, per-engine, and benchmarked. Click a check to mark it verified:

06

Update Cadence & Dependencies

Keeping it alive

Monthly SoV · quarterly prompt-set review
Monthly Re-run the prompt set per engine, update SoV and competitor benchmarks, annotate what moved and why.
Quarterly Review the prompt set itself; add emerging buyer questions, retire dead ones, and note the change so trends stay honest.
On model updates Major engine releases can reshuffle citations; re-baseline SoV soon after so a model-driven swing isn't mistaken for a strategy result.
07

Impact Weightage & Results TAT

What it moves, and how fast

GEO outcome6%

The steering enabler: SoV doesn't earn citations, but it's the metric that directs every other GEO investment to the engine and prompts where you're actually losing share.

SEO outcome3%

Mostly GEO-specific, though the prompt set and competitive framing inform classic content and keyword priorities too.

Editorial estimate of this node's contribution to your total GEO / SEO outcome. Nodes overlap, so weights don't sum to 100.

Results TAT 2–4 weeks Confirming signalA per-engine, benchmarked SoV report ships and starts driving where effort goes; the first cycle usually exposes a hidden per-engine gap.

Tools for this

Go deeper from Measurement

Always current

These links resolve live: what you get is generated or filtered the moment you click.

Sources