Choose the event and universe

Share of voice is meaningless until you define what is being shared. You might measure the share of confirmed brand mentions among selected competitors, the share of answers naming a business, or the share of cited domains. These are different metrics. Name the event, competitor list, prompt set, surfaces, and period. Avoid presenting a sample-based measure as total market share. If a report covers only your selected questions, put that limitation beside the chart. A transparent metric can still be useful for comparing repeated observations within a stable research design.

Decide how multiple brands count

An answer can mention several products. Decide whether each receives one event, whether repeated mentions are deduplicated, and whether unrelated names are excluded. Keep the rules consistent. If you calculate brand-event share, the denominator is all qualifying brand events, not the number of answers. If you calculate answer presence, several brands can each have high rates and their percentages need not sum to 100. Explain this difference to stakeholders before showing the visualization, because an apparently familiar pie chart can suggest a level of exclusivity that the underlying answers do not have.

Use a worked calculation

Hypothetical example: ten answers contain eight qualifying competitor events: four for Acme, three for Beacon, and one for Cedar. Acme has 50% of those events. If its four events occur across four answers, its answer-presence rate is 40%. These are both correct under their definitions. Neither means Acme owns half of all AI discovery. Save the raw event table and inspect ambiguous names. The calculation is simple; the difficult work is ensuring the observations represent the intended businesses and that the sampled questions remain relevant to actual buyer needs.

Keep the comparison set stable

Adding a competitor changes the denominator even when existing answers remain unchanged. Adding branded prompts can also raise a company's presence without any real improvement in broad discovery. Maintain a fixed comparison set for trend reporting, and show expanded sets separately. Note mergers, renames, and product discontinuations. Bing's own citation-share terminology should retain its official meaning rather than being blended with a custom prompt metric. Label platform measures and in-house calculations distinctly, and document any transformation so readers can understand why two seemingly similar shares may differ on the same date.

Interpret share alongside usefulness

A share increase can come from irrelevant mentions or factual mistakes. Pair the chart with recommendation relevance, product accuracy, and landing-page outcomes where observable. Inspect examples before declaring a competitive win. Use the measure to identify questions worth investigating, not to justify arbitrary content volume. If your share drops because the test now includes an unrelated segment, fix the measurement design rather than publishing off-topic pages. The best share-of-voice report makes the numerator and denominator inspectable and connects movement to a specific buyer task that the business has a legitimate reason to serve.

PUT IT TO WORK

Define the denominator and competitor set before using share-of-voice metrics in AI search reporting.

Sources & further reading

Official documentation informs this guide. Examples are illustrative, and product interfaces or limits can change.

THE WORKSPACE WE'RE BUILDING

From search signals to a useful first draft.

SERP.so is in development. Join the waitlist for a workspace that connects research, content review, and publishing.

Join early access ↗