How to compare your brand's AI share of voice.

The slow way: six tabs, twenty prompts, one stale matrix

The manual version: open six tabs — ChatGPT, Perplexity, Google AI Mode, Gemini, Claude, Copilot — and type the same prompt into each. Screenshot, highlight the brand mentions, tally into a spreadsheet: prompts in rows, brands in columns, ones and zeros. Twenty prompts across six engines is a full day. The result: cited in 11 of 20 on ChatGPT, 7 of 20 on Perplexity, 3 of 20 on Gemini.

Is that good? You have no idea — there is no baseline and no trend. So next week you do it again, realize the recurring cost is twelve hours a month, and stop. The matrix ages out before the next board meeting. The first scan was never the problem; the recurring scan is, and share of voice without a trend line is a screenshot, not a metric.

The eCommerce Insights way

1. Add your brand and the top three competitors. Names plus storefront URLs; public catalog data attributes product citations to the right rival. Start with whoever the first scan shows actually sharing your answers.
2. Select the prompt set. 30–60 category-typical buyer prompts, suggested from your category and hero SKUs, fully editable. Narrow prompts measure your terrain precisely; broad prompts measure category reach. Most brands run both views.
3. Run the scan. Every prompt against all six engines, every brand mention and product citation logged. Brand mentions and product citations are tracked separately — the ratio between them is itself diagnostic.
4. Read the dashboard. Your share per engine and overall, beside each competitor, on a weekly trend line. Drill into any engine for the specific prompts won, lost, and tied — and into any movement for the SKUs behind it. That drill-down is what separates this from brand-level tools; the argument is laid out in brand monitoring vs SKU tracking.
5. Export the two-slide summary. Slide one: share per engine with the quarter trend. Slide two: competitor movement with likely drivers ("Brand X gained 8 points on Perplexity following their March PDP rewrite"). The ROI pairing for the same deck is prove AI search ROI to leadership.

How to calculate AI share of voice

The formula is the same one media teams have used for decades, pointed at a new denominator:

share of voice = your brand's mentions ÷ total category mentions × 100, per engine

In practice: fix a prompt set, run it against one engine, and count two things in the answers — mentions of your brand or products (the numerator) and mentions of every brand in the category including yours (the denominator). Divide, multiply by 100, and repeat per engine, because the engines retrieve from different indexes and the per-engine spread is usually the headline finding. If you want to measure share of voice by hand this week, twenty prompts on one engine is enough for a first read — just run the set more than once, since AI answers vary between sessions, and freeze the prompts before you start so next month's number is comparable.

What "good" looks like

Directional reads from eCommerce Insights tracking as of early 2026 (illustrative):

Below 15% brand share usually signals entity-recognition gaps; a brand that dominates Perplexity but vanishes on ChatGPT has an engine-specific indexing issue worth its own diagnosis. Lots of brand mentions with few product citations means the gap is PDP-level — the most fixable shape.

Related jobs