You audit 50 prompts across ChatGPT, Perplexity, and Gemini. You count brand mentions. Your competitor gets 35. You get 12. What does that actually mean?
It means their share of voice in your category is roughly three times yours. And that single number tells you more about your competitive position than any traffic report.
Quick answer: what is share of voice in AI visibility?
Share of voice (SOV) measures the percentage of mentions captured by a brand across a tracked set of prompts, divided by the total mentions in the category. If you run 100 prompt-LLM combinations and your brand appears in 30 responses while competitors collect 70, your SOV is 30%. It's the AI-era equivalent of market share, but built from actual LLM responses, not paid ad spend or media coverage. SOV makes the most sense at the category level (tracked across 20-50 prompts that define your market) and per platform (ChatGPT SOV often differs sharply from Claude SOV).
How share of voice is calculated
The math is simple, the rigor is in the prompt selection.
- Define your prompt set. 20-50 prompts that represent the questions your ideal customer would ask. Mix category prompts ("best CRM for freelancers"), comparison prompts ("[your brand] vs [competitor]"), and use-case prompts ("how to do X").
- Run each prompt across all 7 LLMs. ChatGPT, Perplexity, Gemini, Claude, Grok, Copilot, Google AI Overview. Use fresh sessions to avoid history bias.
- Count brand mentions. For each prompt-LLM combination, note which brands appear. A brand mentioned once per response counts once.
- Aggregate. Total your brand's mentions, total category mentions, divide. That's your SOV.
If you tracked 30 prompts across 7 LLMs (210 total checks) and your brand was named 63 times while all category brands together were named 210 times, your SOV is 30%.
SOV vs other competitive metrics
Share of voice isn't the only way to measure your competitive position. Here's how it relates to other AI visibility metrics:
| Metric | What it measures | When to use |
|---|---|---|
| Share of voice | Your % of total category mentions | Compare your reach against the full competitive set |
| Mention rate | % of your tracked prompts where you appear | Measure your category coverage |
| Position in answer | First, middle, or last mention | Read recommendation strength |
| Sentiment | Positive, neutral, negative tone | Read how AI describes you |
| Citation rate | % of mentions backed by a source link (Perplexity) | Measure authority signal strength |
SOV is the headline metric for competitive benchmarking. The others give the nuance that explains why SOV is what it is.
Traditional SOV vs AI SOV
The term "share of voice" is borrowed from advertising and PR. The mechanic is different in AI:
| Dimension | Traditional SOV | AI SOV |
|---|---|---|
| Data source | Paid ad spend, media impressions, SERP clicks | LLM responses to tracked prompts |
| Auction or editorial? | Yes (paid auction, editorial choice) | No (model decides what to mention) |
| Update frequency | Monthly or quarterly | Daily (LLMs shift in days) |
| Cost lever | Budget increase | Authority + content + mentions |
| Reach | Audience exposed to ad / article | Anyone asking a relevant prompt |
The takeaway: classic SOV maps to spending and pitching. AI SOV maps to the GEO discipline: clear positioning, depth content, third-party citations, structured answers.
Benchmarks: what's a healthy AI share of voice?
Reference points observed on Mentionable accounts (May 2026):
- Category leaders: 40-60% SOV on their target prompts, consistent across 3+ LLMs.
- Strong challengers: 20-35% SOV, growing 3-5 points per quarter.
- Niche players: 10-15% SOV on a narrow prompt set, but high position-in-answer (often first mentioned).
- Invisible brands: under 5% SOV despite an active business. Usually a positioning or third-party mention problem.
Absolute SOV numbers matter less than the trend. A brand at 25% SOV gaining 4 points each quarter is in a healthier competitive position than a brand stuck at 40% with no growth.
How to improve your share of voice
Three levers, ordered by impact:
1. Target the prompts you're losing. Run a competitor visibility audit to identify the prompts where competitors are mentioned and you're not. For each, build content that directly addresses the query with depth and clear positioning.
2. Strengthen the sources LLMs cite. Perplexity shows its sources explicitly. Check what domains it cites in your category, then earn placements on those sites: guest posts, comparison articles, niche directories. Citation-worthy mentions on authoritative sites lift SOV across all LLMs that reference them.
3. Fix the structural extraction signals. Schema markup, FAQ sections, answer-first formatting, clear E-E-A-T signals. These don't directly raise SOV but they make your content more likely to be extracted when an LLM searches the web in real time.
SOV tracking in practice
Manual SOV tracking works for a one-time check but breaks down quickly. Running 30 prompts across 7 LLMs every week is 210 checks per cycle. With 5 competitors to track, that's effectively impossible to maintain by hand.
Automated tools like Mentionable run the prompt set daily across all 7 LLMs and compute SOV by brand, by platform, and over time. The data you actually need: trend lines per competitor, alerts when a competitor's SOV jumps, and breakdowns showing which prompts moved.
Related concepts
Share of voice is one face of AI visibility. To get a complete read, combine it with AI mentions (the raw count), AI citations (the source-level data), and LLM brand tracking (the temporal dimension). For the full GEO discipline that drives SOV up, see GEO and LLMO.
