AI share of voice shows how often, in what position and on what grounds a brand appears in AI answers compared with its competitors, across a defined set of decision questions. Classic share of voice measures the amount of visibility; in AI answers, framing matters as much as amount. Of two brands mentioned equally often, one may be described as the “preferred option” and the other as an “alternative”. Meaningful measurement is done on the decision questions where the brand actually competes, and ends with the cause of the gap, not a ranking.
An industrial packaging manufacturer sees that it’s mentioned in AI answers almost as often as its competitor. The numbers are equal. Reading the answers changes the picture: the competitor is described as “a BRC-certified manufacturer widely chosen in the food industry”. The brand appears in one line under “other companies that could be considered as alternatives”.
Equal count, unequal standing. Classic share of voice measurement can’t see this difference. Competitor analysis in AI answers requires reading how the answer is built, not just the numbers.
Why does share of voice still matter?
Share of voice rests on one of the most robust findings in marketing effectiveness research. Les Binet and Peter Field’s analyses of hundreds of campaigns in the IPA databank showed that brands with a share of voice above their market share (excess share of voice) tend to gain market share over time. The commonly cited rule: every 10 points of excess share of voice is associated with roughly 0.5 points of annual market share growth.
That relationship is about which brands buyers recall when they think of the category. In the AI era, buyers don’t have to recall brands; AI hands them a list. According to G2’s 2026 research, 69% of B2B software buyers chose a different vendor than originally planned based on AI chatbot guidance, and a third bought from a vendor they had never heard of before. The new stage for share of voice is the AI answer.
Why does share of voice need a new definition?
Classic share of voice measured the share a brand held of total category visibility in advertising, press or social media. The unit was impressions, stories or mentions.
AI answers are different. The answer is limited to a few brands, they’re ordered, and each comes with a reason. Buyers read the answer like a recommendation. So what needs measuring is how the brand is mentioned as much as how often.
There’s also the variability problem. In SparkToro and Gumshoe.ai’s January 2026 research, the chance of two answers to the same question containing the same brand list was below 1 in 100; the same order, about 1 in 1,000. Position in a single answer isn’t reliable for competitor comparison. What’s reliable is each brand’s rate of entering the answer across many runs.
Four dimensions of AI share of voice
| Dimension | Question | How to read it |
|---|---|---|
| Appearance share | In how many answers across the decision question set do the brand and competitors appear? | Each brand’s appearance rate across repeated runs |
| Position | Is the brand recommended first, mid-list or among “others”? | As a band, not an exact rank, together with its stability across runs |
| Framing | On what grounds and with which attributes is the brand described? | Classified into roles such as preferred option, specialist, alternative, budget option |
| Source | Which sources does the answer rely on for each brand? | Reveals the source types where the competitor is strong |
Read together, the four dimensions show why a competitor is ahead. A competitor leading on appearance share usually appears in more external sources. A competitor leading on framing usually publishes proof of condition more clearly.
A worked example: appearance share
Suppose a brand measures a set of 20 decision questions, running each question five times in three tools: 300 answers in total. The brand appears in 90 answers, competitor A in 150, competitor B in 60. With 300 total brand mentions across answers, appearance shares are: brand 30%, competitor A 50%, competitor B 20%.
That number is a starting point, not a conclusion. Run the same calculation by persona and the picture can change: the brand may beat competitor A in the procurement manager’s questions while being absent from the quality manager’s. Share with the persona holding veto power matters more than the average.
Which questions should it be measured on?
The most common mistake in competitor analysis is measuring on generic category questions. In “best packaging manufacturers”, big brands always lead. That reflects market size, not sales opportunity. The share of voice reported by AI visibility tools on the market (such as Semrush’s AI modules, Ahrefs Brand Radar or Profound) is mostly based on these category-level prompts; it’s useful for tracking trends but may not show the questions the brand actually wins and loses.
Meaningful competitor analysis is done on the decision questions where the brand actually competes: the conditions of situations where sales meets the competitor, reasons for lost bids, products customers switch from. The decision question map collects these questions systematically. We describe the full measurement method in how to measure AI visibility.
How is the competitor set defined?
The competitor set in AI answers doesn’t always match the brand’s own competitor list. A company the sales team has never met may constantly appear next to the brand in AI answers. Or the reverse: the strongest competitor in the field may be weak in AI answers.
So the competitor set comes from two sources: competitors the sales team knows, and the companies mentioned most often alongside the brand in the simulation. The second group often carries a new insight: which category, and which shelf, the brand sits on in the buyer’s eyes.
Reading why a competitor is ahead
The purpose of competitor analysis isn’t a ranking table; it’s finding the cause of the gap. For the competitor leading in each decision question, ask: On what grounds is the competitor recommended? Which source does that rest on? Does the brand have the same information; if so, why doesn’t it carry into the answer?
The answers usually point to a Communication, Content or Authority GAP. A competitor’s strength often lies not in its product but in publishing its information more clearly, or in appearing more consistently in external sources.
Limits of the measurement
AI share of voice is a periodic, relative measure. Answers change across runs, models update, competitors add content. One period’s numbers shouldn’t be read as a definitive ranking; they should be tracked as a trend. Measuring across several tools prevents one model’s bias from determining the result.
How does Recro run competitor analysis?
The Recro Insight Model simulates the competitor set on the brand’s decision questions and reports the four dimensions together. For the leading competitor in each question, the cause of the gap and the action needed to close it are identified.
One simulation · 4 key pillars · 90+ steps
What sets Recro apart is four pillars built from scratch for each brand. Together they form a single simulation of 90+ steps, with each pillar producing the input for the next.
- Persona builder: Decision-maker profiles that represent the brand’s real buyers, specific to its sector and sales structure.
- Question builder: The brand-specific decision questions these people ask AI along their buying journey.
- Report builder: A report that reads the answers against the brand’s goals and competitors, together with the sources behind them.
- Action recommendation builder: A prioritized list of actions that closes the sector-specific signal gaps.
- The persona and question builders move competitor analysis from category questions to the brand’s real competitive questions. Share of voice is calculated separately for each persona.
- The report builder puts the four dimensions side by side: appearance share, position band, framing and the links that source the conversation. KPIs are set against the brand’s own goals: for one brand the target may be the top band, for another getting onto the list in the veto holder’s questions.
- The action recommendation builder works with a competitor content–equivalent recommendation approach: in every question where a competitor leads, the content and sources behind it are examined, and the equivalent content and signal the brand needs to respond in the same question are defined.
Sample finding format · illustrative
- Question set: Screening · Quality manager · Food packaging
- Appearance: The brand and the competitor appear at similar rates.
- Framing: Competitor: “certified preferred option”; brand: “alternative”.
- Reason: The competitor publishes its certification scope and audit history as text; the brand shows only a logo.
- Equivalent: Write certification scope, audit frequency and food-contact compliance on product pages.
Executive summary
- Share of voice is one of the most robust leading indicators of market share growth; the AI answer is its new stage.
- Share of voice in AI answers isn’t only amount; it’s read with position, framing and source, by persona, on the questions where the brand actually competes.
- The goal is finding the cause, not a ranking. A competitor’s strength often comes from publishing its information more clearly.
Frequently asked questions
Does AI share of voice replace classic share of voice?
No, it complements it. Classic measurement shows media and advertising visibility; AI share of voice shows recommendation visibility at the moment of decision. They answer different questions.
How many companies should the competitor set include?
Usually three to six: competitors the sales team meets often and the companies mentioned most often alongside the brand in the simulation.
How do we measure framing?
The reason sentences in answers are classified: preferred option, specialist, alternative, budget option and so on. It’s a qualitative measure; with a consistent classification scheme, it becomes comparable across periods.
Is the analysis meaningful if the competitor is much bigger than us?
Yes. Big competitors lead in generic questions, but in conditional decision questions expertise can outweigh volume. The analysis shows in which questions the brand can compete with the big competitor.
How often should results be updated?
Continuous monitoring with monthly reviews is enough for most B2B brands. An interim measurement can follow a competitor’s major launch or a big content change.
In which questions, and on what grounds, are your competitors ahead in AI answers? Recro builds competitor analysis on decision questions specific to your brand and shows the cause of every gap.
Request a demo insight report →
Sources
- IPA · Binet & Field, The Long and the Short of It
- Demand Gen Report · Half of B2B software buyers now start their research with AI chatbots: G2 (2026)
- Search Engine Land · AI recommendation lists repeat less than 1% of the time (SparkToro, Gumshoe.ai)
This article was prepared with AI assistance and published under the review of Mehmet Semih İpek.
