AI share of voice is a metric that measures how frequently and prominently your brand appears in the responses generated by answer engines like ChatGPT, Gemini, and Perplexity, relative to your competitors. It matters because an AI answer is a synthesized answer, not a ranked list, and either your brand is in it or it isn't.
How it is calculated
Semrush gives the basic formula: take the AI mentions of your brand, divide by total AI mentions across all brands in your category, and multiply by 100. So if your brand earns 10 mentions and the total across every brand in your category is 100, your AI share of voice is 10%.
Tools count it in different ways. Some count answers instead of mentions: the percentage of AI-generated responses in your category that mention your brand. Others weight position. Semrush's Brand Performance report calculates it from both how many times your brand is mentioned and how high it appears. Pick one method and keep it. Numbers from two methods do not compare.
How it differs from classic share of voice
Classic share of voice measures visibility in paid ads, organic search, or social media. The AI version looks only at how answer engines respond to industry-relevant prompts.
The inputs differ too. Answer engines draw on editorial content, third-party reviews, and structured data, so your share of voice reflects how credibly and consistently your brand is represented across the web, not only what you publish. That is why the work behind it overlaps with E-E-A-T and entity SEO: engines name brands they can identify and have reason to trust.
Why each engine gives a different number
Every AI system has its own training data and method of retrieving and presenting sources from the live web. OptimizeGEO notes that Perplexity weights recency heavily, so its citation patterns tend to shift faster than on other platforms, and that Gemini draws from Google's index and YouTube. So report the number per engine first, then the total.
An example
Say you sell a time tracking app for freelancers. You write down the questions a buyer asks before choosing one, such as "which time tracker should a freelancer use". You ask each one on several AI engines and list every brand each answer names. Then you count.
HubSpot gives the arithmetic: if answer engines reference brands 100 times across a defined set of prompts and your brand accounts for 25 of those mentions, your AI share of voice stands at 25%. The rest belongs to the brands buyers hear instead. That list may not match your Google rivals, because citation frequency reflects content structure, entity authority, and freshness.
How to measure it
- Fix a prompt set. Use buyer questions in buyer words, with no brand names in them.
- Ask each prompt on each engine more than once. AI answers change between runs, which is why a visibility run repeats every question.
- Record every brand named in each answer, yours and the others.
- Work out the share per engine, then overall.
- Repeat on the same prompt set. Tracking it over time reveals whether your brand is gaining or losing ground.
No source we found sets a fixed schedule. Measure again after you ship changes and the engines have had time to recrawl the pages. Daily checks mostly measure noise.
If you build with Claude Code or Codex, Shipfound's visibility run asks 10 buyer questions on ChatGPT, Perplexity, Gemini, Claude and Google's AI Mode and AI Overviews, 4 runs each, and counts how often you are named.
Common mistakes
- Prompts with your brand name in them. They inflate your share and tell you nothing about buyers.
- Mixing methods. A mention count and an answer count give different numbers for the same answers.
- Counting a cited link as a mention. If the answer cites your page but recommends someone else, the buyer heard someone else's name. Track citations separately, as an AI citation count.
- One run on one engine. That is a sample of one.
Frequently asked questions
What is a good AI share of voice percentage?
There is no shared benchmark across categories. It depends on how many brands compete in your answers. Compare the number with your own last measurement and with the brands named next to you.
What does 100% share of voice mean?
Every brand mention in your prompt set was yours. That is rare, and it usually means the prompts are too narrow or contain your brand name.
What does 50% share of voice mean?
Half of all brand mentions across your prompts went to you, and the other half went to every other brand combined.
How do I track AI share of voice?
Keep the same prompts, engines and method, and measure again after each round of changes. Measuring it gives you a benchmark and shows whether your generative engine optimization efforts are working.