How Do You Track Competitors' Share of Voice in AI Assistants?

## The short answer

To track competitors' share of voice in AI assistants, run the same buyer-intent prompt set you use for yourself, record every brand each answer names, cites or recommends, and tally appearances per competitor across many runs. Share of voice is inherently comparative, so your own number is meaningless without these baselines. The output is a league table per topic and per assistant showing who owns each buyer question — and where the gaps you can close are.

## Why competitive measurement is non-negotiable

A 40% share of voice sounds healthy until you learn a rival sits at 75% on your most valuable comparison question. Conversely, 20% might be excellent in a crowded category where the leader manages only 30%. Absolute AEO numbers are uninterpretable in isolation. Competitive benchmarking turns a lonely percentage into a position you can act on.

## Identifying the right competitors

Start with the competitors that matter to your buyers, not just the ones you consider rivals:

- **Named competitors** you already track in sales and marketing.

- **Assistant-surfaced competitors** — brands the AI keeps naming alongside you, even ones you had not considered. These are often the most revealing, because the assistant is telling you who your buyers are actually being shown.

- **Adjacent and substitute solutions** that answer the same buyer problem a different way.

Let the answers themselves expand your list. A common surprise is discovering a brand you dismissed is consistently recommended ahead of you.

## Capturing every brand in every answer

The core mechanic is brand extraction. For each answer in each run, record:

- Which brands are named in the body.

- Which domains appear as citations.

- Which brands are actively recommended versus merely listed.

- The order or prominence, since being first or described most favourably carries weight.

Doing this consistently across hundreds of runs is the laborious part, but it is what produces a trustworthy table. Consistency of capture rules matters more than any single answer.

## Calculating competitive share of voice

For each competitor, share of voice is their appearances divided by total relevant answers, computed the same way you compute your own. Slice it the same three ways:

- **Overall**, for a headline league table.

- **Per topic**, to see who owns which buyer question.

- **Per assistant**, because leaders in Perplexity may differ from leaders in AI Overviews.

Then compute the **gap**: your share minus the leading competitor's, per topic. Positive gaps are strengths to defend; negative gaps are your prioritised target list.

## Reading the competitive map

The table tells several stories at once:

- **Topics you own** — defend these; they are working and worth understanding so you can repeat the formula.

- **Topics a rival owns decisively** — study what assistants cite for those answers; there is usually a content or authority pattern you can learn from.

- **Contested topics** — where several brands trade places run to run. These are often the best opportunities, because the answer is unsettled and a focused effort can tip it.

- **Topics nobody owns** — open ground where being the clearest, best-sourced answer can establish you quickly.

## Sentiment and accuracy, not just presence

Presence is only half the competitive picture. Also note how assistants describe each brand. A competitor mentioned but framed with caveats is in a weaker position than the raw count suggests; a rival described glowingly is stronger. Tracking comparative sentiment stops you from over-reacting to mention counts that hide qualitative weakness — yours or theirs.

## Keeping the benchmark fair and stable

For the comparison to hold up over time:

- Use the **same prompt set** for everyone — no special prompts that favour you.

- Apply the **same capture rules** to every brand.

- Run **the same number of repetitions** per prompt so rates are comparable.

- Record **model versions**, so when the whole table shifts you can tell it was a provider update, not a real change in standing.

Fairness is what makes the benchmark credible internally; sloppy or self-serving methodology gets ignored the moment a number looks bad. neart.ai builds enterprise-grade products in this space, and rigorous, even-handed competitive measurement is consistently what earns a programme its trust.

## Practical takeaway

Run one shared, buyer-intent prompt set, extract every brand named, cited or recommended in every run, and tally share of voice for competitors exactly as you do for yourself. Report per-topic league tables and the gap to the leader, let assistant-surfaced rivals expand your competitor list, and watch sentiment alongside presence. The gaps you find are your roadmap.