Skip to main content
How-to

How to Measure Share of Voice in AI Search

By · Updated · 7 min read

What Measuring Share of Voice Actually Involves

Measuring share of voice for an ecommerce store means building a repeatable process: choosing a fixed set of real buyer questions, running them against multiple AI systems, and calculating what percentage of the resulting answers name your brand versus competitors. Unlike a single citation check, this is not a one-time task.

It's a monthly practice that only produces a meaningful number once you've run it more than once on the same query set, so you can see whether your position is improving, stable, or losing ground.

Step 1. Build a Representative Set of Category Queries

Start with 15 to 30 real buyer-intent questions specific to your product category. Not generic keywords like "best protein powder," but actual questions a shopper would type or speak to an AI assistant: "what's the best protein powder for someone on keto" or "how much biltong should I order for a party of ten."

Pull these from your own customer support tickets, product reviews, and search terms already showing up in your site search, since those reflect real language your buyers use rather than guesses. A keyword idea generator can help round out the list. Keep the list stable once built. Changing the query set between measurement periods breaks the comparison you're trying to build.

Step 2. Run Every Query Across the Major AI Engines

Run each question individually against ChatGPT, Claude, Perplexity, and Google AI Overviews, or Gemini, depending on which conversational AI systems matter most to your buyers. Ask each question fresh, in a new conversation, without prior context that could bias the answer toward your brand.

Screenshot or copy the full text of every answer rather than skimming, since citations sometimes appear in a source list separate from the visible prose, and you'll need the complete response to log accurately in the next step.

Run the full query set in a single sitting if possible, rather than spreading it across several days. AI models are updated frequently, and answers to the same question can shift week to week for reasons unrelated to your content, so a measurement round completed close together in time is a cleaner comparison point than one stretched over weeks.

Step 3. Log Every Source Cited in Every Answer

For each answer, record every brand or source named, not just whether your brand appeared. A spreadsheet with one row per query and one column per engine, containing every brand mentioned, is enough for a small query set.

Note ties, multiple brands cited in a single answer, as separate entries rather than picking just one, since AI answers frequently name two or more sources rather than a single winner. This raw log is what makes the following calculation step defensible instead of a guess.

Step 4. Calculate Your Share of Voice Percentage

Count the total number of citation events across your full log, every brand-naming instance across every query and engine, then count how many of those events named your brand. Divide your count by the total to get your share-of-voice percentage for that query set.

Do the same for your two or three biggest competitors so you have a comparative baseline, not just an isolated number. A 12 percent share of voice means little on its own. A 12 percent share against a competitor's 40 percent tells you exactly where you stand.

Consider calculating the percentage twice: once across your whole query set for a category-level view, and again broken down by subtopic if your questions span more than one clear theme. A blended number can hide a category where you're strong and another where you're nearly invisible.

Step 5. Repeat Monthly and Track the Trend

Rerun the identical query set against the identical engines on a fixed monthly schedule, and log each round in the same spreadsheet so you can chart the percentage over time. A single measurement is a snapshot. Three or four consecutive months is a trend.

A trend is what tells you whether new content, schema fixes, or a broader AI-visibility push is actually working. If your query list or catalog grows large enough that manual monthly runs become unsustainable, a dedicated AI-visibility monitoring tool can take over the same process without changing the underlying method.

Frequently asked questions

How many buyer questions do I need for a reliable share of voice measurement?

15 to 30 real, specific buyer questions is a practical range for most ecommerce categories. Fewer than 10 makes the percentage noisy and easily swung by a single answer. More than 30 becomes hard to sustain manually without a dedicated monitoring tool.

Which AI engines should I include when measuring share of voice?

At minimum, include ChatGPT, Perplexity, and Google AI Overviews, since together they cover a large share of how buyers now get AI-generated answers. Add Claude or Gemini if your buyers are known to use them, but consistency across measurement periods matters more than the exact engine list.

Do I need to log every brand mentioned, or just whether my brand appears?

Log every brand mentioned in every answer. Recording only your own brand's presence or absence makes it impossible to calculate a real percentage, since share of voice requires dividing your citation count by the total citation events across every brand named in the query set.

How often should I repeat the measurement?

Monthly is a sustainable cadence for most stores and is frequent enough to see a real trend within a quarter. Running it less often than quarterly makes it hard to connect a specific content change to a specific shift in the number.

What do I do once I have my share of voice percentage?

Identify the specific queries where competitors are cited and you're not, check whether your site has content directly answering those questions, and publish or rewrite against the clearest gaps. Then rerun the same query set on your next scheduled measurement to see whether the new content moved the number.

MG
Written by

Matt is the founder of RunOctopus. He built All Angles Creatures from zero to page-1 rankings in reptile feeder insects using exactly this method. Turning a hard, entrenched niche into RunOctopus's proof store for programmatic SEO and AI search citation.

Connect on LinkedIn →

See what Ollie would build for your store

Free architecture preview. No card required. Five minutes.

Generate Preview →