Steps in this guide
Step 1
Define your competitor set
Identify the 3-5 products that appear most often alongside yours in category recommendations. These are your share-of-voice comparators. Run a few manual queries in ChatGPT to confirm which names keep appearing in your category.
Step 2
Define your query set
Choose 10-20 category discovery queries that represent how buyers find products like yours. These should be non-branded 'what is the best tool for X' style questions, not brand searches.
Step 3
Run queries and record all brand mentions
For each query, run it 3 times and record every brand mentioned - including competitors. Note which position each brand appears in the recommendation list. For share-of-voice calculation, you need total brand appearances, not just yours.
Step 4
Calculate raw share of voice
Share of voice = (your brand mentions / total brand mentions for all tracked brands) x 100. If your brand appears 24 times out of 120 total brand appearances across all queries and runs, your SoV is 20%.
Step 5
Calculate per-LLM share of voice
Run the same calculation separately for each AI assistant. Your SoV in Perplexity may be very different from your SoV in ChatGPT. Per-LLM breakdown tells you where to focus your improvement efforts.
Step 6
Track SoV over time
Record your SoV at a fixed weekly interval. The trend matters more than any single data point. A SoV that is 18% this week but was 12% a month ago signals that your AEO efforts are working. A flat or declining SoV signals something needs to change.
Skip the manual version of this guide
AskRank runs this exact workflow automatically across ChatGPT, Claude, Perplexity, and Gemini, on a schedule.
AI share of voice tells you how much of the recommendation space your brand occupies relative to your competitors. A 40% share of voice means your brand appears in 40% of the total brand mentions across your tracked queries and AI assistants.
This metric is more meaningful than raw mention count because it contextualizes your performance. Being mentioned 30 times sounds good until you realize your main competitor is mentioned 120 times.
Why share of voice beats simple mention tracking
Traditional brand monitoring counts mentions. Share of voice compares those mentions to the total competitive landscape.
Consider two scenarios:
- Scenario A: Your brand is mentioned in 60% of ChatGPT runs for your target queries
- Scenario B: Your competitors collectively appear in 90% of runs, while you appear in 60%
In Scenario A, 60% sounds healthy. In Scenario B, you are clearly trailing a competitor who shows up nearly every time.
Share of voice gives you Scenario B’s context by default.
Step 1: Define your competitor set
AI share of voice is only meaningful relative to the competitors that appear in the same recommendation space as you.
Start by running 5-10 of your target queries in ChatGPT or Perplexity. Note which brand names appear. The ones that show up repeatedly across multiple queries are your true AI competitors - they may not be the same as your pricing-page competitors.
Limit your tracked competitor set to 3-5 brands. Tracking more becomes unwieldy manually and makes the SoV metric harder to interpret.
Step 2: Build your query set
The quality of your SoV measurement depends entirely on the quality of your query set. Use queries that:
- Represent actual buyer intent (not what you wish people searched)
- Cover multiple use cases within your category
- Include “best X for Y” patterns, “alternatives to Z” patterns, and “how do I” patterns
A well-designed query set of 20 questions gives you a much more reliable SoV signal than 50 random queries.
Step 3: The SoV calculation
The formula is straightforward:
SoV = (your brand appearances / total brand appearances) x 100
Where “total brand appearances” is the sum of appearances for all brands you are tracking, across all queries and all runs.
Example:
- Query set: 20 queries
- Runs per query: 3
- Total query runs: 60
- Your brand appearances: 28
- Competitor A appearances: 42
- Competitor B appearances: 19
- Competitor C appearances: 11
- Total appearances: 100
- Your SoV: 28%
A few important notes on the calculation:
- Count each appearance once, regardless of whether the brand appears first, second, or third in a list
- If your brand appears twice in a single response (mentioned in the list and then discussed separately), count it once
- A brand that is not mentioned at all in a run counts as 0 for that run
Per-LLM share of voice
The aggregate SoV across all AI assistants is useful, but per-LLM SoV is where the actionable insights live.
Your brand might have 35% SoV in Perplexity but only 15% in ChatGPT. This tells you different things:
- High Perplexity SoV + low ChatGPT SoV: your review site presence (which Perplexity crawls) is stronger than your training-data presence. Focus on content and authority building to improve ChatGPT.
- Low Perplexity SoV + high ChatGPT SoV: your content has historically been well-represented in training data, but your real-time web presence is weaker. Focus on getting featured on sites Perplexity cites.
AskRank tracks per-LLM visibility score and competitor comparison automatically across ChatGPT, Perplexity, Claude, and Gemini. The competitor bar chart on the dashboard shows you SoV visually at a glance.
What is a good share of voice?
There is no universal benchmark - it depends entirely on your competitive landscape.
A general framework:
- SoV above 30% in a 4-competitor space: you are the leader or close to it
- SoV 15-30%: competitive presence, room to grow
- SoV under 15%: you are being underrepresented relative to your actual market position
If you have strong customer reviews, a polished product, and good SEO, but AI SoV under 15%, that is a signal worth investigating. There is a specific gap between how your product is perceived online and how AI assistants are interpreting those signals.
Tracking SoV over time
Calculate SoV weekly using the same queries, the same competitor set, and the same number of runs per query. Consistency in methodology matters more than frequency.
Set a baseline in week 1. Revisit every 4 weeks to identify trends. Week-over-week changes are noisy. Month-over-month trends are where meaningful signals live.
See also: AI share of voice in the glossary, How to track competitors in AI answers, and How to improve your AEO.