AI share of voice is the percentage of brand mentions in AI assistant answers that belong to your brand, measured across a fixed set of buyer prompts and assistants such as ChatGPT, Gemini and Perplexity. It tells you how often you come up, compared with competitors, when people ask AI about your category.
Share of voice has been a marketing metric for decades. AI assistants add a new place where the conversation about your category happens, and it needs its own measurement. This guide covers how to calculate it, what distorts it, and how to tell whether your number is good.
Key takeaways
- AI share of voice is your brand’s mentions divided by all brand mentions across a fixed set of prompts and assistants.
- The prompt set defines the metric. Change the prompts and you change the number, so keep them stable.
- Measure per assistant. A blended score can hide a strong position in one and an absence in another.
- There is no universal “good” score. Judge yours against your direct competitors and against your own trend.
- Pair share of voice with citation data to understand why the number moves.
What AI share of voice measures
At its simplest:
AI share of voice = mentions of your brand ÷ mentions of all tracked brands, over a defined set of prompts, assistants and time period.
If you track five CRM brands across 20 prompts, run each prompt in three assistants, and count every time any of the five brands is named, your share is the fraction of those mentions that were yours. You’ll also hear it called LLM share of voice; it’s the same idea.
It’s related to, but different from, a few other numbers you may see:
- Mention rate (or presence rate): the percentage of answers that include your brand at all. Useful, but it ignores competitors.
- Citation share: the fraction of cited URLs that point to your domain. This measures whether you’re a source, not whether you’re recommended.
Share of voice is the one that best reflects competitive standing, because it’s relative by design.
How to measure AI share of voice, step by step
1. Fix your competitor set
Pick the 3 to 8 brands you actually compete with for the same buyer. Include the category leader even if you don’t consider them a direct rival, because assistants often will.
2. Fix your prompt set
Choose prompts that reflect real buying questions, from broad category queries to specific use cases. Write them down and don’t edit them casually. If you add or reword prompts, note the date so you can explain any shift. See choosing prompts to track for how to build a balanced set.
3. Run each prompt repeatedly in each assistant
AI answers aren’t deterministic. Run each prompt multiple times per period in each assistant you care about.
4. Count mentions consistently
Decide on rules and stick to them:
- Count a brand once per answer, or count every mention? Once per answer is simpler and less sensitive to long, repetitive responses.
- Do negative mentions count? They usually should, but read the answer text so you know how you’re framed.
5. Calculate per assistant, then overall
Compute share of voice in ChatGPT, Gemini, Perplexity, Claude and others separately before you blend them. Then decide whether the overall number should weight assistants equally or by how much your audience uses each one.
6. Track the trend
Re-run on a regular cadence and plot the result. The line matters more than any single point.
What distorts the number
A few common mistakes make share of voice look better or worse than it is:
- Branded prompts. “Is HubSpot a good CRM?” will always mention HubSpot. Keep your own brand name out of the core set, or report branded prompts separately.
- Too few runs. With one run per prompt, a single lucky answer moves the number a lot.
- A drifting prompt list. Adding easy prompts inflates the score without anything changing in the market.
- Ignoring absent competitors. If a new entrant starts appearing and isn’t in your tracked set, your share looks stable while your real position slips.
What’s a good AI share of voice score?
There isn’t a benchmark that applies across categories, and you should be wary of anyone quoting one. A score depends on how many competitors you track, how crowded the category is, and how broad your prompts are. Five tracked brands and twenty tracked brands produce very different ranges.
Better ways to judge your number:
- Against your competitors. Are you ahead of, level with or behind the brands you lose deals to? Your rank among tracked brands is often more telling than the raw percentage.
- Against your market position. If you’re a leader in revenue or customers but trail in AI share of voice, assistants are underrepresenting you, and there’s likely a source problem to fix.
- Against yourself. A rising trend after you’ve improved pages or earned new coverage is the clearest signal that your work is landing.
- Per prompt type. You might lead on use-case prompts but disappear on broad “best X” prompts. That pattern tells you where to focus.
Turning share of voice into action
The number tells you where you stand. Citations tell you why. When your share drops on a prompt, look at which sources were cited in the answers where you were absent. Often the same handful of review sites or comparison articles keep appearing. Getting onto those pages, or making sure they describe you accurately, is usually the most direct lever. Our guide on how AI assistants choose the sources they cite explains what tends to get a page selected.
A monthly share of voice review
- Prompt set unchanged, or changes dated and noted
- Share of voice calculated per assistant
- Rank among tracked competitors noted
- Untracked brands appearing in answers reviewed
- Biggest gains and losses by prompt identified
- Top cited domains for lost prompts listed
- One or two actions chosen for next month
Measure AI share of voice automatically with Citegram
Calculating this by hand gets heavy fast: prompts multiplied by assistants multiplied by runs. Citegram is a Chrome extension that does the counting for you:
- Add your prompt set. Use the fixed list from step 2.
- Run it in the real assistant apps. Citegram asks each prompt in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews in background tabs, using the accounts already signed in to your browser. No API key is involved, and questions count toward your own plan usage.
- Check the per-answer detail. Each answer is stored with its text, whether and where your brand is mentioned, and the cited sources in rank order.
- Read share of voice per brand and per assistant. This gives you the per-assistant breakdown from step 5 without a spreadsheet, plus a list of the most cited domains.
- Re-run to build the trend. Repeated runs show how your share of voice in ChatGPT and the other assistants moves over time. Runs can be stopped and resumed, and nothing is lost if Chrome closes.
If you track several assistants, this guide on tracking your brand across AI assistants covers how their results tend to differ.
FAQ
How do you calculate AI share of voice?
Divide your brand’s mentions by the total mentions of all tracked brands across the same prompts, assistants and time period. Count each brand once per answer to keep long responses from skewing the result. Calculate it per assistant first, then blend if you need a single number.
What is a good AI share of voice?
There’s no universal benchmark. A good score is one that puts you ahead of the competitors you lose deals to and trends upward over time. Compare within your own tracked set rather than against figures from other categories.
Why is my share of voice different in each assistant?
Each assistant retrieves and weighs sources differently, and some rely more on live web search than others. Different sources lead to different shortlists, which is why per-assistant reporting matters.
See your own share of voice
You can start measuring with a single prompt on the free plan, then move to Pro at $10 a month if you want to track a full prompt set. Check the plans and see how your brand compares.
ChatGPT
