Classic share of voice asked how much of the search results page you owned. In AI answers the question changes: when an assistant names a few options, how often is one of them you?
Daniel Arons · Jun 2026 · 6 min read
Marketers have tracked share of voice for decades. In paid media it meant your slice of total ad spend in a category. In SEO it meant the percentage of valuable keywords where you ranked, weighted by position and search volume. The logic was always the same: a category has a finite amount of attention, your competitors are fighting for it, and your share is the portion you capture. The metric mattered because it predicted market share. Owning more of the conversation tended to translate into owning more of the customers.
AI assistants break the old measurement, then quietly rebuild it in a new shape. When someone asks ChatGPT, Claude, Gemini, or Perplexity for the best option in your category, there is no page of ten blue links to divide up. There is one answer, and it usually names a short list. Either you are on that list or you are not. The question is no longer how much of the results page you own. It is how often, across the questions buyers actually ask, the assistant recommends you instead of a competitor. That is share of voice for the AI era, and it is becoming the single number that tells a brand where it really stands.
From share of voice to share of recommendation
It helps to rename the thing so you measure it correctly. Classic share of voice counted impressions, rankings, and listings. The unit was visibility on a page the buyer still had to read and judge for themselves. AI assistants compress that judging step. They do not just show you ten options to sort out. They narrow the field and hand back a recommendation, often with reasons attached. So the modern version of the metric is better described as share of recommendation: of all the times an assistant answers a buying question in your category, what fraction of those answers name you as a choice worth considering.
The distinction is not pedantic. A recommendation carries weight a ranking never did. When a person scans a results page, they apply their own skepticism to every listing. When an assistant says you are one of the strong options for a need, it has already done the filtering on the buyer's behalf, and the buyer tends to trust the shortlist. That is why a few points of share of recommendation can matter more than a much larger swing in classic search visibility.
“On a results page the buyer does the filtering. In an AI answer the assistant filters first, then hands the buyer a shortlist. Being on that shortlist is closer to a referral than an impression.”
How AI share of voice differs from the SEO version
If you already track SEO share of voice, it is tempting to assume the AI version is the same calculation pointed at a new surface. It is not, and the differences change how you measure.
There is no fixed results page to divide
SEO share of voice rests on a stable, observable artifact: the ranked list of results for a query. You can scrape it, see who appears in which position, and compute everyone's share. AI answers have no such fixed list. Two people can ask the same question and get differently worded responses that name a different mix of brands. The answer is generated, not retrieved from a fixed index, so your share has to be estimated across many runs rather than read off a single page.
Position is fuzzy, framing is everything
In SEO, rank one is unambiguously better than rank eight, and the value of each slot is well understood. In an AI answer, being named first is not always the prize. What matters is how you are framed. An assistant might list you first but describe you as the budget option, or name you third while calling you the best fit for enterprise teams. Share of recommendation has to account for not just whether you appear, but in what context and with what sentiment, because that framing is what shapes the buyer's next move.
The question set is the battlefield, not the keyword list
SEO organizes the world around keywords. AI buyers ask full, conversational questions, often layered with constraints: best tool for a small team on a tight budget that integrates with a specific system. Your share of voice depends on how you perform across that realistic set of natural-language questions, not across a list of head terms. This is why the AI version sits on top of SEO rather than replacing it. Strong SEO still feeds the sources assistants draw from, but the unit of measurement moves from keyword rankings to question outcomes.
How to measure your share of voice across a prompt set
Because there is no page to scrape, measuring AI share of voice means running a deliberate experiment and counting the results. The method is straightforward once you accept that you are sampling a generated system, not reading a fixed one.
Build a representative prompt set
Start with the questions real buyers ask on the way to choosing in your category. Include broad ones (what is the best option for this job), comparison ones (this brand versus that brand), and constrained ones that carry the qualifiers your ideal customers care about. The prompt set is the foundation of the whole measurement, so it should reflect actual buying intent rather than the questions you wish people asked. A few dozen well-chosen questions tells you far more than hundreds of generic ones.
Run each question repeatedly and across assistants
Ask each question multiple times, on each assistant your buyers use, because the same prompt produces variation. Running it once tells you what happened once. Running it many times tells you the pattern, which is what you actually want to manage. Record every brand named in every response, not just yours, so you can see the full competitive field.
Turn the responses into a share number
Now do the counting. For a given question set, your share of voice is the proportion of responses that name you out of all the brand mentions, or out of all the responses, depending on which view you want. Layer in the context: how often you appear first, how often you are framed positively, and which competitors crowd you out when you are absent. The result is a defensible figure you can track over time and break down by question theme and by assistant. Our guide to measuring AI visibility goes deeper on building this into a repeatable process.
Why it is the core AI-visibility KPI
Plenty of metrics orbit AI visibility, but share of voice deserves to sit at the center for a simple reason: it is comparative, and buying is comparative. A raw count of how often you are mentioned can look healthy while a competitor is mentioned twice as often for the same questions. Presence tells you whether you are in the room at all. Share of voice tells you how much of the room is recommending you instead of someone else, the part that actually maps to who wins the deal.
It is also the metric that survives the volatility of generated answers. Any single response is noisy and can change from one run to the next. Share of voice, measured across a stable prompt set and many runs, smooths that noise into a trend you can act on. When you ship new content, earn a mention on a source the models trust, or correct a misconception, you watch your share move. That feedback loop is what turns AI visibility from a thing you worry about into a thing you manage.
“Presence tells you whether you are in the room. Share of voice tells you how much of the room is recommending you instead of a competitor. Only one of those maps to who wins the deal.”
How to grow your share of voice
Once you can measure share of voice, growing it stops being guesswork and becomes a targeted exercise. The measurement itself shows you where to push.
Fix absence before you chase framing
Start with the questions where you never appear at all. Those zeros are the largest, cheapest gains, because moving from absent to present is a bigger swing than improving how you are described once you are already named. Look at what assistants cite when they answer those questions well for your competitors, and work on becoming a credible source for the same need. The pages, mentions, and structured information the models draw on still have to exist and have to be clear.
Shape how you are described where you do appear
For the questions where you show up but get framed weakly, the work shifts to clarity. If an assistant calls you the cheap option when you compete on outcomes, the source material it reads is probably leading with price. Make the positioning you want the easy, well-supported story to tell about you. Our walkthrough on increasing your share of AI recommendations covers the moves that tend to shift framing.
Treat it as a continuous loop, not a campaign
AI answers keep changing as models update and the underlying sources shift. A share of voice number from last quarter is a starting line, not a finish line. Re-run the prompt set on a regular cadence, watch how your share moves against named competitors, and let the gaps direct the next round of work. The brands that win here are the ones that make this a standing rhythm rather than a one-time audit.
Where Aethon fits
Share of voice in AI search is measurable, comparative, and movable, but only if you run it as a disciplined process rather than the occasional manual spot check. That is the work Aethon is built for. We map how a defined set of buyer questions gets answered across the major assistants, track who gets recommended and how they are framed, and turn that into a share of voice figure you can watch over time and tie back to the actions that move it. To see what your category looks like through this lens, take a look at how Aethon works or book a demo and we will walk through your numbers together.
Frequently asked questions
What is share of voice in AI search?
It is your share of the recommendations AI assistants give in your category. Across the buying questions people ask ChatGPT, Claude, Gemini, and Perplexity, it measures how often you are named as an option compared with everyone else who gets named. Because assistants hand back a shortlist rather than a results page, it is often called share of recommendation.
How is AI share of voice different from SEO share of voice?
SEO share of voice divides up a fixed, observable results page by ranking and search volume. AI answers are generated, so there is no fixed page to divide. You estimate your share across many runs of natural-language questions, and you account for framing and context, not just whether you appear and in what position.
How do I measure my share of voice across AI assistants?
Build a prompt set that reflects how buyers actually ask, run each question repeatedly on every assistant your buyers use, and record every brand named in every response. Then count the proportion of responses that name you out of all mentions, and break it down by question theme, by assistant, and by how you are framed.
Why is share of voice the most important AI visibility metric?
Because buying is comparative. A raw mention count can look fine while a competitor is named twice as often for the same questions. Presence tells you whether you appear at all; share of voice tells you how much of the recommendation space you own against rivals, which is what actually maps to who wins the deal.
How do I grow my share of voice in AI answers?
Start with the questions where you never appear, since moving from absent to present is the biggest swing. Then improve how you are framed where you already show up by making your desired positioning the easy, well-supported story to tell. Re-run the measurement on a regular cadence and let the gaps guide each round of work.
Written by
Daniel Arons
Co-founder & CEO, Aethon AI
Daniel co-founded Aethon AI in November 2025 to close the gap between how marketers measure AI visibility and what AI is actually doing with their brands. Before Aethon, he spent eight years building digital marketing programs in New York across SaaS, financial services, and consumer brands. He holds an MPA from Baruch College and a BA in Public Relations from SUNY Oswego.