How to measure AI share of voice against your competitors (2026)

The quick answer

AI share of voice (SOV) measures the percentage of AI-cited sources for a query set that point to your brand versus competitors, counted per engine and per country. Compute AI share of voice in five steps: build a fixed query set, run each query across ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews, count brand citations versus total competitor citations, weight each query by importance, then track the delta weekly. As of 2026, the metric only stays comparable when the query set, engines and country are frozen between runs. This playbook gives the formula, a worked example with real numbers, and a tracking cadence.

Note on method: AI engines change their cited sources frequently. Freeze your query set, engine list and country before measuring, and re-run on a fixed schedule so each AI SOV number is comparable to the last.

What AI share of voice measures and why it differs from SEO share of voice

AI share of voice quantifies how often AI engines cite your brand as a source, expressed as a percentage of all brand-attributable citations across a defined query set. SEO share of voice counts ranking positions on Google's ten blue links; AI share of voice counts citations inside generated answers, where only a handful of sources appear and position matters less. The two metrics diverge because an AI engine can cite a page that ranks #7 on Google and ignore the #1 result.

Three properties separate AI share of voice from classic SEO share of voice:

• Unit of measurement: AI share of voice counts citations (a source named or linked in an answer), not ranking positions.
• Multi-engine surface: a single query produces different cited sources across ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews, so AI share of voice is computed per engine then aggregated.
• Per-country variance: AI engines cite different sources for the same query in the US versus the UK, so AI share of voice without a fixed country is not comparable.

For the formal definition, see the glossary entry on AI share of voice and the broader concept of AI visibility. This playbook owns the how-to-measure methodology, not the definition.

Step 1: define a fixed query set that matches buyer intent

A query set is the frozen list of 20 to 50 questions you measure AI share of voice against, chosen to mirror how buyers actually prompt AI engines. A query set of 30 buyer-intent questions produces a stable AI SOV signal; a set under 10 queries swings too much per run to trust. Lock the query set before the first measurement, because changing the questions between runs makes the resulting AI SOV numbers non-comparable.

Build the query set from three sources:

1. Category questions: "best CRM for B2B SaaS", "top GEO tools 2026", "what is the best project management software". These map to commercial intent and surface competitor sources directly.
2. Problem questions: "how to track AI citations", "how to reduce churn in SaaS". These surface the educational sources AI engines trust.
3. Comparison questions: "X vs Y", "alternatives to X". These reveal which brands AI engines name side by side.

A worked anchor: a B2B CRM vendor might lock 30 queries, including "best CRM for B2B SaaS". You can preview which sources an engine cites for a query like that on the best CRM for B2B SaaS answer page before committing it to your set. Tag each query with a weight from 1 to 3 by commercial value, which Step 4 uses.

Step 2: run every query across each engine and country

Each measurement run executes the full query set against every target AI engine, in a single fixed country, and records the cited sources returned for each query. Run all 30 queries across 5 engines in 1 country to produce 150 answer-source lists per run. Hold the country constant (for example, United States) across runs, because an AI engine cites different sources for "best CRM" in the US versus Germany, and mixing countries corrupts the AI share of voice trend.

Capture three fields for every cited source:

Doing this by hand across five engines is slow and drifts as prompts change. An AI-citation API returns the cited sources for any query, per engine and per country, as structured JSON, which removes manual copy-paste and keeps runs identical. See the API reference for the response shape, or the build-it-yourself path in how to build an AI visibility monitor on the citation API.

Step 3: count brand citations versus competitor citations

Citation counting attributes every cited source in the run to your brand, to a named competitor, or to a neutral source, then sums the totals per engine. Count 1 citation each time an answer cites a domain you own; count 1 competitor citation each time an answer cites a domain on your competitor list. A source that belongs to neither your brand nor a tracked competitor (a review site, a forum, a publisher) counts as neutral and stays out of the SOV denominator unless you measure share of total citations instead.

Define the attribution rules before counting, and keep them identical every run:

• Brand citation: the answer cites a domain you own (your marketing site, docs, or owned subdomains).
• Competitor citation: the answer cites a domain on a pre-declared competitor list (lock the list to 3 to 8 named rivals).
• Neutral citation: the answer cites any other domain (G2, Reddit, a news site). A page multiple engines cite at once is a hotspot source, and winning placement on one shifts AI share of voice fastest.

A practical decision: most teams measure AI share of voice as brand-vs-competitor (denominator excludes neutral sources), because it answers "who is winning this category in AI answers". Share of total citations (denominator includes neutral sources) answers a different question and produces a smaller percentage.

Step 4: compute the weighted AI share of voice percentage

Weighted AI share of voice divides your weighted brand citations by the total weighted citations across your brand and competitors, expressed as a percentage. The formula is: AI SOV = (sum of brand citations times query weight) divided by (sum of all brand-plus-competitor citations times query weight), multiplied by 100. Weighting by query value prevents a low-value query from distorting the score as much as a high-value buyer query.

The formula, stated plainly:

AI SOV (%) = ( Σ (brand_citations × query_weight) ÷ Σ (brand_and_competitor_citations × query_weight) ) × 100

Worked example for a B2B CRM vendor, one engine, one country, three queries from a 30-query set:

Weighted AI share of voice = (8 / 28) times 100 = 28.6% for that engine and country. Repeat per engine, then aggregate with an engine weight if some engines matter more to your audience (for example, weight Perplexity higher if your buyers use it most). To compute AI SOV in code instead of a spreadsheet, the same formula drives the build-an-AI-visibility-monitor playbook.

Step 5: track the delta over time and segment by engine

AI share of voice becomes useful only as a tracked time series: a single 28.6% reading means little, while a move from 28.6% to 34.0% over four weeks proves a GEO action worked. Re-run the frozen query set on a fixed cadence (weekly is the common default in 2026), keep the engines, country and competitor list identical, and record the AI SOV delta per engine. A delta you can attribute to a specific content change is the metric that justifies GEO spend.

Three segmentation cuts make the trend actionable:

• Per engine: a brand can hold 40% AI share of voice on Perplexity and 18% on Google AI Overviews, which points effort at the weaker engine.
• Per query cluster: AI SOV on comparison queries versus problem queries shows where competitors out-cite you.
• Per country: AI SOV in the US versus the UK reveals market-specific gaps, since AI engines cite different sources by locale.

A common pitfall: AI engines refresh cited sources on their own schedule, so a one-week dip can be engine noise rather than a real loss. Tracking a four-week moving average separates signal from noise. Many teams cross-reference AI SOV with AI referral traffic in GA4 to confirm that rising citations convert to visits.

Common measurement mistakes that break AI share of voice comparability

Most AI share of voice numbers are wrong because the measurement setup changed between runs, not because the brand's visibility changed. AI SOV is a relative metric, so any drift in the query set, engine list, country or competitor list silently corrupts the trend. Lock all four variables before the first run and document them, because a comparison across two different setups measures the setup, not the brand.

The five mistakes that most often invalidate an AI SOV trend:

1. Changing the query set mid-track: adding or removing queries shifts the denominator and breaks comparability.
2. Mixing countries in one number: AI engines cite different sources per country, so a blended-country AI SOV is meaningless.
3. Counting one engine as the whole: ChatGPT alone is not AI share of voice; a per-engine breakdown is required.
4. Ignoring neutral sources entirely: if a hotspot review site dominates an answer, brand-vs-competitor SOV can look healthy while total citation share is low.
5. Comparing to SEO rankings: a top Google ranking does not equal an AI citation, a divergence covered in Google rankings vs AI citations.

Measure AI share of voice with Getspotted

Getspotted is the AI-citation API plus MCP server: one /search call returns the sources ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews cite for any query, in a chosen country, with the cross-engine hotspots and the contacts behind each source. The API returns cited domains as structured JSON, which is the exact input the five-step AI share of voice method needs, so you can automate the count, the weighting and the weekly delta instead of copy-pasting from five chat windows.

Developers and agencies building AI SOV tracking start with the docs; for a deeper how-to on outranking competitors as sources, see how to get cited in AI answers.

FAQ

How do you calculate AI share of voice?

Calculate AI share of voice by dividing weighted brand citations by total weighted brand-plus-competitor citations across a fixed query set, then multiplying by 100. Count citations per engine and per country, weight each query by commercial value, and re-run the same query set on a fixed schedule so each AI SOV percentage is comparable.

What is a good AI share of voice percentage?

A good AI share of voice depends on the number of tracked competitors: against 4 competitors, an even split is 20%, so anything above 20% means you out-cite the average rival. As of 2026 there is no universal benchmark, because AI SOV is relative to your competitor list and query set; the meaningful target is a rising delta over time, not an absolute number.

How many queries do you need to measure AI share of voice?

Use 20 to 50 buyer-intent queries to measure AI share of voice reliably. A set of 30 queries produces a stable signal, while a set under 10 swings too much per run to trust. Lock the query set before the first measurement, because changing queries between runs makes the AI SOV numbers non-comparable.

Should AI share of voice be measured per engine or combined?

Measure AI share of voice per engine first, then aggregate with optional engine weights. A brand can hold 40% AI share of voice on Perplexity and 18% on Google AI Overviews, and a single combined number hides that gap. Per-engine AI SOV points your GEO effort at the weakest engine.

How often should you measure AI share of voice?

Measure AI share of voice weekly in 2026, using a frozen query set, engine list and country each run. AI engines refresh cited sources on their own schedule, so a weekly cadence with a four-week moving average separates a real visibility change from engine noise.

What tools measure AI share of voice in 2026?

GEO dashboards such as Profound, Otterly and Peec report AI share of voice as a built-in chart, while an AI-citation API like Getspotted returns the raw cited sources so you compute AI SOV in your own stack (capabilities as of 2026, verify on each vendor's docs). Choose a dashboard for turnkey reporting and an API when the AI SOV data must live in your product, covered in how to choose a GEO tool.

Does AI share of voice include sources that are not my brand or a competitor?

Brand-versus-competitor AI share of voice excludes neutral sources (review sites, forums, publishers) from the denominator, which answers "who wins this category in AI answers". Share of total citations includes neutral sources and answers a different question, producing a smaller percentage. Pick one definition and keep it identical across every run.