Home › Guides › AI share of voice vs competitors

AI share of voice: measure your brand vs competitors across ChatGPT, Gemini & Claude

When a buyer asks an AI "what's the best CRM for a small agency?", the model names a handful of brands. AI share of voice answers the obvious next question: of all the brands AI assistants name in your category, what slice is you — and what slice is each competitor? It's the single number that turns "are we visible in AI search?" into a scoreboard you can benchmark and move.

What is AI share of voice?

AI share of voice (SoV) is the percentage of relevant AI answers that mention your brand, relative to all brand mentions across a defined set of prompts. Classic search "share of voice" measured your slice of paid or organic visibility on a results page. AI SoV measures your slice of the brands an assistant volunteers when a buyer asks a question — across ChatGPT, Gemini and Claude. If the model never names you, you don't exist at the moment of decision, no matter how good your website is.

The share-of-voice formula

The math is deliberately simple so it stays comparable over time and across engines:

• AI Share of Voice (%) = your brand mentions ÷ total brand mentions × 100, counted across your whole prompt set and all samples.
• The denominator is every mention of every brand in your competitive set — not the number of prompts. One answer can mention three brands; all three count.
• Compute it per engine (ChatGPT vs Gemini vs Claude) and blended, because the same brand can dominate one assistant and be invisible on another.

Worked example: across 20 buyer-intent prompts you and four competitors are mentioned a combined 140 times. You account for 35 of those mentions. Your AI share of voice is 35 ÷ 140 = 25% — and if the category leader sits at 40%, you now know the exact gap to close.

The non-determinism trap

Ask the same question three times and you can get three different brand lists. A tool that asks each prompt once reports a coin flip as if it were the truth — and your SoV can swing 10–15 points between two single-shot runs for no real reason. Reliable measurement samples each prompt 3–5 times per engine and reports a mention rate plus a stability score. Stability tells you whether your 25% is a solid, repeatable 25% or a noisy average of 10% and 40%. Without it, you can't tell a real movement from sampling noise — which is the difference between a vanity dashboard and decision-grade data.

How to pick your competitive set and prompt set

Your SoV is only as honest as the inputs. Two choices decide everything:

• Competitive set — name 3–6 brands a buyer would realistically weigh against you, not aspirational giants. The set defines the denominator, so keep it stable between runs to make the trend meaningful.
• Prompt set — write buyer-intent questions, the ones people actually ask before purchasing: "best X for Y," "X vs Z," "alternatives to Z," "cheapest X for small teams." Avoid prompts that name your brand (that biases the answer) and avoid pure-trivia prompts that no buyer asks.

Aim for 15–30 prompts spanning the buying journey. That's enough to average out noise without ballooning your cost per run.

Benchmark your % vs named competitors

foXLabs publishes GEO Competitive Benchmark on Apify — a pay-as-you-go actor that computes AI share of voice the right way:

• Runs your prompt set through ChatGPT, Gemini and Claude with live web search — no API keys to wire up.
• Samples each prompt 3–5× and reports a mention rate + stability, so your SoV is repeatable rather than a single-shot guess.
• Returns a ranked share-of-voice table — your % beside each named competitor, per engine and blended.
• Surfaces the gap to the leader and the prompts where you're absent, so you know exactly where to compete next.

From scoreboard to strategy

Once you know your SoV, the levers are concrete: target the prompts where competitors are named and you aren't, earn mentions and citations on the third-party domains those answers already trust (review sites, comparison articles, authoritative guides), and re-measure to confirm the trend is moving up. To go from what's my share to why, pair the benchmark with AI Brand Monitor, which breaks down the underlying mention rate, sentiment and citation gap behind your SoV %.

How often should you measure?

AI answers drift as models and the web change, so SoV is a monitoring discipline, not a one-off. Schedule a weekly run with a fixed competitive and prompt set, watch the delta, and wire a webhook to Slack so a drop in your share alerts your team automatically.

Frequently asked questions

What is AI share of voice?

The percentage of relevant AI answers that mention your brand relative to all brand mentions across a defined prompt set. If you and four competitors are named a combined 100 times and you are named 30, your AI share of voice is 30%.

How do you measure share of voice in ChatGPT?

Pick a competitive set and buyer-intent prompts, run each prompt through ChatGPT (and Gemini and Claude) several times, count each brand's mentions, then divide your mentions by total brand mentions. Because answers vary run to run, report a mention rate plus a stability score, not a single shot.

How many times should you sample each prompt?

Sample each prompt 3–5 times per engine. AI answers are non-deterministic, so one query is a coin flip; 3–5 samples give you a reliable mention rate and a stability score so you know how much to trust each number.