Measurement
AI share of voice: how to measure it without fooling yourself
AI share of voice is the percentage of brand mentions your company earns in AI answers to a fixed set of buyer prompts, compared with the competitors you track. It is the closest thing AI search has to market share, and it is also the easiest number in AEO to get wrong. This guide shows the formula, a worked example, how many runs make the figure trustworthy, and how to report it so a CFO can rely on it.
Get a free AEO visibility check
Two lines is enough. A specialist replies within one business day.
The short answer
What is AI share of voice?
AI share of voice is your brand's mentions divided by all tracked brands' mentions across a fixed prompt set, times 100, measured per engine. Because AI answers change between runs, each prompt must be sampled several times, and the result should be read as a monthly trend, never as a single snapshot.
Quotable points
- Formula: your mentions divided by all tracked brands' mentions, times 100, per engine.
- Run each prompt several times; one run per prompt produces noise, not a metric.
- Report mention share and citation share separately; they measure different things.
- Always state the scope: prompts, engines, competitors, runs and dates.
How do you calculate AI share of voice?
Divide the number of times your brand is mentioned by the total mentions of every brand you track, across the same prompts and engines, then multiply by 100. That is the standard formula, and Ahrefs, Nightwatch and most tracking vendors use a version of it.
The formula is the easy part. The number only means something when four inputs stay fixed from month to month: the prompt set, the list of competitors, the engines, and the number of runs per prompt. Change any of them and the percentage moves even if nothing changed in the market.
Some tools weight the result by how much each engine is actually used. Ahrefs' October 2026 guide to AI share of voice explains that Brand Radar can weight impressions by platform-adjusted search volume, using ratios of 0.115 for ChatGPT, 0.011 for Gemini and 0.01 for Perplexity. Weighting is useful for an executive summary, but keep the unweighted per-engine figures too, because your buyers may lean heavily on one engine.
What does a worked example look like?
Here is an illustrative example for a B2B SaaS company selling payroll software with three competitors. The numbers are invented to show the arithmetic, not taken from any real brand.
The team writes 40 buyer prompts such as "best payroll software for a 50-person company" and runs each prompt 5 times in ChatGPT, giving 200 answers. Across those answers, the four tracked brands are mentioned 480 times in total. The company is mentioned 96 times, so its ChatGPT share of voice is 96 divided by 480, or 20%.
Two further numbers make the result useful. The company appeared in 78 of the 200 answers, a visibility rate of 39%. And its own pages were linked as a source in 22 answers, which is its citation rate. A brand can have a healthy mention share but a weak citation share, which usually means engines know the name from third-party sources while its own pages are not being used.
| Brand | Mentions in 200 ChatGPT answers | Share of voice | Answers mentioning brand |
|---|---|---|---|
| Competitor A | 168 | 35% | 121 (61%) |
| Competitor B | 132 | 28% | 97 (49%) |
| Your company | 96 | 20% | 78 (39%) |
| Competitor C | 84 | 17% | 66 (33%) |
Why does one run per prompt give a misleading number?
Because AI answers are not stable. SparkToro and Gumshoe.ai ran 2,961 prompts 60 to 100 times each across ChatGPT, Claude and Google's AI, and Search Engine Journal reported in January 2026 that the same list of brands came back less than 1% of the time, and the same list in the same order less than 0.1% of the time.
The same study found that how often a brand appears across many runs was far more consistent than its position in any single answer. That is the strongest argument for share of voice as a frequency metric, and against any tool or agency report that quotes an "AI ranking position".
Engine settings add more variance. Semrush found that only 25% of cited sources overlapped between ChatGPT's different reasoning modes, so the same prompt can draw on different evidence depending on how the answer is generated. In practice we sample each prompt at least three to five times per engine per month and read three-month trends, not single reports.
Is mention share the same as citation share?
No. Mention share counts how often your brand name appears in answers. Citation share counts how often your own URLs are linked as sources. Both matter, and they move for different reasons.
Mention share rises when the sources engines trust describe you: reviews, comparison articles, Reddit threads, analyst lists and press. Citation share rises when your own pages are crawlable, indexed and answer the question better than anything else. Digital PR moves the first; answer-first content and AI crawler access move the second.
There is also a first-party version of citation share. Bing Webmaster Tools added Citation Share and Compare views to its AI Performance report in June 2026, covering citations in Copilot and Bing's AI summaries. It covers only Microsoft's surfaces, but unlike any third-party tool it comes from the engine itself, so we use it as a check on tool data.
Which prompts should go into the set?
Prompts that a real buyer types before choosing a vendor. A set built from your own keywords will flatter you; a set built from buyer questions shows where you actually lose.
Pull candidates from sales call notes, support tickets, Search Console queries, Reddit and community threads, and the questions your sales team hears on demos. Then group them by stage, so the report can show where share is won or lost, not just a single blended score.
Keep branded prompts out of the share of voice calculation. A question like "is your company any good" will mention you every time and inflate the number. Track branded prompts separately for accuracy, which is where wrong AI answers about pricing or features show up.
- Category discovery: "best X for Y", "top X tools".
- Comparison: "X vs Y", "alternatives to X". These map to your comparison pages.
- Problem: "how do I fix Z", where a vendor might be suggested.
- Local or regional variants if you sell in several countries.
- 25 to 100 prompts in total is enough for most brands; consistency matters more than size.
What is a good AI share of voice?
There is no universal benchmark, and anyone quoting one without a scope is guessing. The fair comparison is against your competitors on the same prompt set and against your own previous months.
A simple rule of thumb is to compare share of voice with your share of the market. If you hold roughly a fifth of the customers in your category but only a tenth of AI mentions, AI answers are underweighting you, and that gap is the business case for AI search optimization. If your share of voice is already above your market share, defending it is cheaper than growing it.
Also look at spread across engines. A brand at 30% in ChatGPT and 5% in Perplexity has an engine-specific problem, often crawler access or source mix, which Perplexity SEO work can address directly.
Should you track AI share of voice manually, with a tool or with an agency?
Manual tracking works for a small prompt set; a tool is worth it once you need several engines and runs; a managed service makes sense when nobody internally has time to keep the method consistent. Our comparison of AEO tools covers the main trackers and their trade-offs.
Whichever route you take, keep a written method: the prompt list, competitor list, engines, run count, location and dates. Without it, you cannot tell a real change from a change in method.
| Approach | Best for | Strength | Weakness |
|---|---|---|---|
| Manual runs in a spreadsheet | Under 30 prompts, one or two engines | Free and fully transparent | Slow; few runs per prompt; easy to drift |
| Tracking tool | 50 to 500 prompts across several engines | Many runs, history, competitor views | Opaque sampling; vendors define share differently |
| Managed tracking service | Teams that want conclusions, not dashboards | Fixed method, reconciled with Search Console, Bing and GA4 | Ongoing cost; you depend on the provider's rigour |
| First-party reports (Bing, Search Console) | Everyone, as a cross-check | Data from the engines themselves | No competitor view; limited engine coverage |
What are the most common share of voice reporting mistakes?
Most mistakes come from changing the method and then reading the change as progress. Watch for these before you put the number in a board deck.
- Adding or removing competitors mid-quarter, which mechanically changes everyone's share.
- Mixing branded and unbranded prompts in one figure.
- Reporting a single run as a monthly result.
- Blending engines into one number without showing the per-engine split.
- Comparing figures from two tools that define share of voice differently.
- Treating share of voice as revenue. Pair it with AI referral sessions and conversions in GA4 from chatgpt.com, perplexity.ai, copilot.microsoft.com and gemini.google.com.
When is AI share of voice the wrong KPI?
When buyers in your category do not use AI assistants to choose, or when you are the only realistic option. A niche manufacturer with three known competitors may learn more from AI referral traffic and lead quality than from share of voice.
It also misleads when the brand facts in answers are wrong. A high share built on an outdated price or a discontinued product is a liability, so read accuracy alongside volume. For most B2B software and ecommerce brands, though, it is worth tracking: a G2 survey of 1,076 software buyers in March 2026 found 51% now start research in an AI chatbot more often than in Google.
If you want an outside view, AEO Services Agency, the AEO and GEO practice of Greenwolf Tech Labs, starts with a free AEO visibility check and runs AI visibility tracking on a fixed prompt set, measured before and after every change. Work is fixed-price after that check. No one can promise a citation, but a clean method tells you honestly whether your share is moving.
FAQ
Frequently asked questions
What is the formula for AI share of voice?
Your brand's mentions divided by the total mentions of all tracked brands across the same prompts and engines, multiplied by 100. Calculate it per engine first, then blend if you need a single figure, and keep the prompt set, competitors and run count fixed between reports.
How many times should I run each prompt?
At least three to five times per engine per measurement period. SparkToro's research found AI engines return the same brand list less than 1% of the time, so a single run tells you very little. More runs narrow the noise; trends across three months tell you more than any one report.
Can I measure ChatGPT share of voice for free?
Yes, for a small prompt set. Run each prompt several times in ChatGPT with search on, log every brand mentioned and every cited URL in a spreadsheet, and apply the formula. It is slow but transparent. Add Bing Webmaster Tools' free Citation Share view as a first-party cross-check for Copilot.
What is the difference between AI share of voice and AI visibility?
AI visibility usually means the percentage of answers that mention your brand at all. Share of voice compares your mentions with competitors' mentions on the same prompts. Visibility can rise while share of voice falls if competitors are growing faster.
Why did my AI share of voice drop this month?
First check whether the method changed: new competitors, new prompts, a different tool setting or fewer runs. If the method is unchanged, look at the cited sources for the prompts you lost; a new comparison article, review roundup or Reddit thread often explains the shift.
Do AI share of voice tools all measure the same thing?
No. Vendors differ on whether they count mentions or citations, how many runs they sample, which engines and locations they use, and whether they weight by engine usage. Compare trends within one tool, and avoid comparing raw percentages across tools.
Explore
Find your industry and market
AEO by industry
AEO by country
Keep reading
Related pages
Get started
Find out what AI says about you
Send your website and three questions your buyers ask. Within one business day we reply with how ChatGPT, Google AI Overviews and Perplexity answer them today, who they cite instead of you, and a fixed-price plan.
- Free AEO visibility check
- Fixed quote before any work starts
- Measured on citations and AI referrals, not promises
Prefer email? info@aeoservicesagency.com
Get a free AEO visibility check
Two lines is enough to start.