# AI Share of Voice: How to Measure It in AI Search

> How to calculate share of voice in AI search results: formulas, prompt sampling, Search Console and Bing Citation Share data, and what really raises it.

URL: https://seorecheck.com/blog/ai-search-share-of-voice · Published: 2026-11-13 · [Українська](https://seorecheck.com/ua/blog/ai-search-share-of-voice.md), [Español](https://seorecheck.com/es/blog/ai-search-share-of-voice.md)

**AI share of voice** is the share of AI-generated answers for your category's questions that mention, recommend or cite your brand, compared with your competitors. You measure it by running a fixed set of prompts many times in ChatGPT, [Google AI Mode](https://seorecheck.com/blog/google-ai-mode-optimization), Copilot or Perplexity and counting who shows up. Then you check it against the first-party data that Google and Bing now publish.

The idea comes from classic marketing, where share of voice meant your part of all the advertising or mentions in a market. In AI search, the "market" is the answers an assistant gives. This guide covers the formulas, the data sources and the sampling rules that keep the number honest.

## What is share of voice in AI search?

Share of voice in AI search is a competitive visibility metric. It shows how often your brand appears in AI answers compared with the other brands that appear for the same questions. It's always relative: 30% means that out of all the brand appearances in your sample, about three in ten were yours.

People use the term for three different things, so define which one you track:

- **Mention share:** your brand is named in the answer text, with or without a link.
- **Citation share:** a page from your domain is listed as a source.
- **Recommendation share:** your brand appears in a list of suggested providers or products.

These can split apart. A software review site might be cited constantly while its own brand is never recommended. A local plumber might be recommended by name but cited only through a directory listing. To track whether assistants talk about you at all, start with our method to [track brand mentions in AI search](https://seorecheck.com/blog/track-brand-mentions-ai-search). Share of voice adds the competitor comparison on top.

## How do you calculate AI share of voice?

Count appearances across a sample of AI answers, then divide your count by the total for every brand you track. Use the same prompts, platforms and time window for every brand. Otherwise the shares aren't comparable. Calculate presence rate alongside it, because share alone hides how often nobody from your list appears.

| Metric | Formula | What it tells you |
|---|---|---|
| Presence rate | Answers that mention you ÷ all answers sampled | How often you're in the conversation at all |
| Mention share of voice | Your mentions ÷ mentions of all tracked brands | Your slice of brand visibility vs competitors |
| Citation share of voice | Citations to your domain ÷ citations to all tracked domains | How often your pages are used as sources |
| Recommendation share | Times you appear in a recommended list ÷ all list appearances | Your slice of "who should I buy from" answers |

**Worked example (illustrative numbers):** you track 30 prompts, run each 5 times in two assistants, and get 300 answers. Your brand appears in 72 of them, so your presence rate is 24%. Across those 300 answers, the five brands you track are mentioned 400 times in total, and 80 of those mentions are yours. Your mention share of voice is 20%.

Count each brand at most once per answer. Otherwise one long answer that repeats a name five times will skew the totals.

## Why shouldn't you track your "rank" in AI answers?

Position inside an AI answer is close to random, so a single ranking number misleads. Visibility measured over many repeated runs is far more stable. That's why share of voice should be built on frequency (how often you appear), not on where you appear in a list.

The clearest evidence comes from a [SparkToro and Gumshoe.ai study](https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/). In it, 600 volunteers ran 12 recommendation prompts through ChatGPT, Claude and Google's AI a combined 2,961 times. The study found less than a 1 in 100 chance of getting the same list of brands in two responses. Identical lists in the same order came up roughly once in 1,000 runs. Even the number of recommendations varied, from two or three to ten or more.

The same study found that **visibility percentage across many runs** was much more consistent. Leading headphone brands appeared in about 55–77% of answers. The authors note that broad categories need bigger samples than narrow ones. They also suggest asking tracking vendors for publicly reviewable research before you buy.

## Which data sources show AI share of voice?

No platform shows you a full competitor share of voice. Bing comes closest with a Citation Share metric for your own site. Google reports AI impressions but no queries or competitors. For brand-versus-brand comparisons, you still need your own prompt sampling or a paid tracking tool.

| Source | What it gives you | Competitors? | Queries? |
|---|---|---|---|
| [Google Search Console Generative AI report](https://support.google.com/webmasters/answer/16984139?hl=en) | Impressions of your links in [AI Overviews](https://seorecheck.com/blog/seo-for-ai-search) and AI Mode, by page, country, device and date | No | No |
| [Bing Webmaster Tools AI Performance](https://blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview) | Total citations, average cited pages, sampled grounding queries and page-level citations in Copilot, Bing AI summaries and partners | No | Sampled |
| [Bing Citation Share](https://blogs.bing.com/search/2026/6/New-AI-Visibility-Insights-in-Bing-Webmaster-Tools-Intents-Topics-Citation-Share-Compare/) | Your percentage of all citations for a grounding query, plus Intents, Topics and period Compare | Share only, no domains | Yes |
| Manual prompt log | Mentions, citations and recommendations across any assistant | Yes | Your prompts |
| Paid prompt-tracking tools | Automated version of the prompt log at scale | Yes | Their prompts or yours |

### What does Bing's Citation Share measure?

Bing's Citation Share is the closest official equivalent to share of voice. Microsoft defines it as [the percentage of citations attributed to your site out of all citations shown across all sites for the same grounding query](https://blogs.bing.com/search/2026/6/New-AI-Visibility-Insights-in-Bing-Webmaster-Tools-Intents-Topics-Citation-Share-Compare/). The same post, published in June 2026, says it does not expose competitor domains, doesn't represent traffic share and doesn't score content quality. It's in preview. It tells you how big your slice is for Copilot and Bing's AI summaries, but not who owns the rest. It's still useful for spotting topics where you're cited only now and then. Group the data by Bing's Topics view to see which subject areas need work.

### What does Google's report add?

Google's [Generative AI performance report](https://support.google.com/webmasters/answer/16984139?hl=en) counts an impression when links to your site were shown to a user in an AI feature on Google Search. It shows no query data and no competitors. Use it as your absolute baseline: if impressions rise while your sampled share of voice stays flat, the whole category is getting more AI exposure. For how this fits next to classic rankings and clicks, see [SEO vs GEO performance measurement](https://seorecheck.com/blog/seo-vs-geo-performance-measurement).

## How do you build a prompt set for share of voice?

Write 20–50 prompts that real buyers would ask an assistant about your category. Mix in problem, comparison and local questions, and tag each one by intent. Freeze the set so every month compares like with like. Then run each prompt several times per platform and log every brand and source.

1. **Collect real questions.** Use sales calls, support tickets, site search and Search Console queries. Write them the way people talk to assistants: "What's the best accounting software for a three-person agency?" rather than "accounting software agency".
2. **Cover the funnel.** Include category questions ("what are the options for…"), comparisons ("X vs Y"), problems ("how do I fix…") and local or price-sensitive variants where they matter.
3. **Pick 4–6 competitors** you actually lose deals to, plus a catch-all "other" column.
4. **Fix the conditions.** Use logged-out or clean sessions where possible, the same country and the same date window. Personalization and location change answers.
5. **Run each prompt several times per platform.** The SparkToro researchers ran each prompt 60–100 times. Few small teams can do that by hand, so treat a handful of runs as a rough signal, not a precise figure.
6. **Log per answer:** brands mentioned, whether each was recommended, cited domains and any factual errors about you.
7. **Calculate monthly** with the formulas above, broken down by platform and intent.

If logging by hand gets too slow, compare what automated trackers actually measure in our guide to [AI citation tracking tools](https://seorecheck.com/blog/ai-citation-tracking-tools) before paying for one.

## What is a good AI share of voice?

There's no published benchmark for a "good" AI share of voice. It depends on how many brands compete in your category and how broad your prompts are. Compare yourself with your own previous periods and with your direct competitors on the same prompt set, not with a number from another industry.

Some practical readings:

- **Low presence rate, decent share:** when you appear, you hold your own, but you're missing from many prompts. That usually points to coverage gaps, meaning questions your site doesn't answer.
- **High citation share, low mention share:** assistants use your pages as sources but recommend other brands. Your content explains the category well, but your own offer isn't clearly stated or backed up elsewhere.
- **One platform far behind:** check that platform's crawler access and index first. Copilot depends on Bing's index, for example.

Be wary of any tool that reports a single "AI visibility score" without showing the prompts, run counts and raw answers behind it.

## How do you improve your share of voice in AI answers?

You raise share of voice the same way you earn citations: be crawlable and indexed, answer the questions in your prompt set clearly, and be described consistently across the web. Google says you don't need [special machine-readable files, AI text files or markup](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) to appear in its AI features.

Google's [AI optimization guide](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) explains that AI Overviews and AI Mode use query fan-out. That means running several related searches at once and grounding answers in results from Google's core ranking systems and Search index. Beyond the technical requirements, your site must also be included in Search generative AI features in Search Console. In practice:

1. **Fix eligibility first.** Allow the search and user-triggered AI bots you want to be cited by (see [AI crawler access](https://seorecheck.com/blog/ai-crawler-access-seo)). Avoid snippet controls that block quoting. Make sure key text doesn't appear only after JavaScript runs.
2. **Answer the losing prompts.** For each prompt where competitors appear and you don't, check whether a page on your site answers it directly. If none does, write one. Query fan-out means supporting questions matter too.
3. **State facts plainly.** Put prices (or a link to them), service areas, integrations and differentiators in visible text that an assistant can quote.
4. **Strengthen entity signals.** Use consistent Organization markup with sameAs links, an accurate [Wikidata entry](https://seorecheck.com/blog/entity-seo-ai-search-engines) where your business qualifies, and matching descriptions on the profiles assistants cite.
5. **Re-measure on the same prompt set** after 30–60 days, and read the change against Search Console and Bing trends.

Checking robots.txt rules, snippet directives, JavaScript-only text and entity markup page by page takes time. A [free SEORecheck audit](https://seorecheck.com/request) runs these GEO checks on your site and reports which ones pass. You can see the format in the [sample report](https://seorecheck.com/sample). For a deeper look at what each platform says about picking sources, read [how AI search engines select sources](https://seorecheck.com/blog/how-ai-search-engines-select-sources).

## Key takeaways

- AI share of voice compares your brand's appearances in AI answers with your competitors'. Choose between mentions, citations or recommendations and keep that choice fixed.
- Measure frequency across many prompts and repeated runs. Position within an answer is too unstable to track.
- Bing's Citation Share gives your slice of citations per grounding query, but no competitor names. Google's report gives AI impressions without queries.
- Competitor comparisons still need a frozen prompt set, logged manually or with a tracking tool that shows its raw data.
- There's no universal benchmark. Track the trend against your own baseline and direct rivals.
- Improvements come from eligibility, direct answers and consistent entity data, not from special AI files.

---

## Want to know what to fix on your site?

Get a free SEO audit preview: your score and real issues, with evidence from your pages.

- [Get a free SEO audit preview](https://seorecheck.com/request)
- [See a sample report](https://seorecheck.com/sample)
