Skip to content
Chalk Labs / AI SEO
How-to guide

How to Measure Your Brand's Share of Voice in AI Search

By Rahil Jain, founder of Chalk LabsUpdated How we research

The short answer

To measure AI share of voice, write a fixed set of 30 to 50 buyer prompts, run each one in the AI engines your buyers use, and record which brands are mentioned or cited. Your share is your count divided by all brand counts. Keep the prompts unchanged for at least a month and repeat runs.

  • Vendors define share of voice differently. Ahrefs counts the percentage of AI responses in a topic set that mention or cite your brand versus rivals.
  • Which prompts you track matters more than how often you run them, according to a Profound test of 753 prompts.
  • AI answers are volatile, so one run proves nothing. Repeat runs and compare like with like.
  • A visibility number says nothing about whether the AI describes you correctly, so score accuracy too.

What does AI share of voice actually measure?

AI share of voice measures how often your brand appears in AI answers compared with competitors, for a fixed set of prompts. Ahrefs defines it as the percentage of AI responses within your topic set that mention or cite your brand versus competitors 1.

Two terms matter. A mention is when your brand name appears in the answer text. A citation is when the AI links to your pages as a source 1. Track them separately, because a brand can be named without a link, and linked without being recommended.

For the plain definition, read what is AI share of voice. This page covers how to measure it.

Do it step by step

0 of 8 done0%

  1. Write the one topic you want to own, such as 'invoicing software for small agencies'. List your five closest competitors, plus any marketplaces AI tends to cite in your field.

  2. Split them into three groups: learning, comparing and ready to buy. Write them in plain buyer language. At least a third of the prompts should name no brand.

    Tip: Use the free AI Prompt List Builder to draft a starting list, then edit it by hand.

  3. Save the prompts in a sheet with an ID for each. Do not edit the wording for at least a month. New ideas go in a separate group.

  4. Use logged-out or fresh sessions so memory and custom instructions do not colour results. Note the engine, model name if shown, country and date for every run.

  5. Paste each answer, or just the brands named and links cited, into the sheet. One row per prompt per run. Add a column for whether the description was correct.

  6. Visibility rate is answers naming you divided by total answers. Share of voice is your mentions divided by all brands' mentions. Calculate both for mentions and for citations.

  7. Mark each answer that names you as correct, partly wrong or wrong. List the wrong claims and the source page the AI seems to have used.

  8. Repeat on the same schedule and chart the change by prompt group. Pick the three prompts where rivals appear and you do not, and plan a page for each.

Your ticks are saved in this browser only.

Copy-ready templates

Tracking sheet header (tab-separated)

text
prompt_id	prompt	group	engine	date	run_no	brands_named	you_mentioned	you_cited	cited_urls	accuracy	notes

Set you_mentioned and you_cited to 1 or 0. Use accuracy values: correct, partial, wrong, n/a.

Spreadsheet formulas for the scores

text
Visibility rate:  =SUM(H2:H500)/COUNTA(A2:A500)
Citation rate:    =SUM(I2:I500)/COUNTA(A2:A500)
Share of voice:   =YourMentions/(YourMentions+Comp1+Comp2+Comp3+Comp4)
Accuracy rate:    =COUNTIF(K2:K500,"correct")/COUNTIF(H2:H500,1)

Adjust column letters to match your sheet. YourMentions and each Comp cell hold a brand's total mention count; add one Comp term per competitor you track.

Sample prompt set across the buyer journey

markdown
**Learning**
- What should I look for in [category]?
- How does [category] work for a [size] business?

**Comparing**
- What are the best [category] options for [use case]?
- [Competitor A] vs [Competitor B]: which is better for [use case]?

**Ready to buy**
- How much does [category] cost for [size] business?
- Which [category] provider should I choose in [city]?

Replace the brackets with your real category, use cases and places. Keep at least a third of the set brand-neutral.

What do you need, and how long does it take?

You need a spreadsheet, a list of your top five competitors, and access to the engines you want to test. By hand, 40 prompts run 3 times in 2 engines is about 240 answers, so allow a full day. To start in about three hours, use 15 prompts in one engine. These times are our estimates. A tracking tool cuts later months to roughly an hour.

Decide your engines first. Your buyers may use ChatGPT, Google AI Overviews and AI Mode, Perplexity, Gemini and Copilot, so pick two or three that matter and stick with them.

How do you design a prompt set that gives a fair answer?

The prompt set is the most important choice you make. Profound tested 753 prompts on seven US platforms over 14 days in June 2026. It found that tracking once a day gave a visibility reading within about 2 percentage points of tracking ten times a day, and concluded that which prompts you track matters more than how often you run them 3.

Profound sells tracking and the test was internal, so the exact numbers may not transfer 3. The principle still holds: build the set carefully. Aim for 30 to 50 prompts, split across the buyer journey: learning, comparing and ready to buy.

Write them the way buyers talk, with and without your category name. Include some prompts where no brand is hinted at, such as 'which tool should a 10-person agency use for invoicing'. Our prompt list guide and the free AI Prompt List Builder help you draft it.

How many times should you run each prompt?

Run each prompt at least three times per engine in the first round, in fresh chats, then compare. Ahrefs says AI answers are inherently volatile and that the best way to measure visibility is to track a sample of prompts and see who gets recommended 2. Its study of AI Overviews also found that results are probabilistic and change with each query 4.

More runs reduce noise, but not unlimited noise. In Profound's test, running a prompt ten times a day cut day-to-day wobble in citation share by about 40% but in visibility by only about 11%. For citation share, most of the remaining movement was platform drift, not sampling 3. So pick a steady schedule, such as weekly or monthly, and keep it.

Also note that many ChatGPT answers involve no web search. Ahrefs found only 14.4% of logged-out responses in its study triggered one 5. Your brand can therefore be recommended from the model's memory, which can differ from what live search returns.

How do you score the answers?

Record four facts per answer: were you mentioned, were you cited, which competitors were named, and was the description of you correct. The first two give visibility; the last gives quality.

Seer Interactive argues that visibility alone can mislead because a visibility metric counts a positive and a negative brand mention the same. It proposes three measures: how often you are mentioned (AI Signal Rate), how accurately AI describes you (Answer Accuracy Rate) and conversion among AI-influenced sessions 6. The article gives no formulas, so the sheet below is our own simplification.

To judge accuracy, write a short fact sheet about your brand first: what you sell, price range, locations, who you serve. Mark each answer correct, partly wrong or wrong against it.

What does a worked example look like?

The numbers below are illustrative, not real data. Imagine 40 prompts run three times each in one engine, giving 120 answers. Count how many answers name each brand, then divide your count by the sum of all brands' counts.

Notice the second column in the table: visibility rate uses a different denominator. Report both, and say which is which, because tools use different formulas and the numbers are not interchangeable.

Illustrative only: 120 answers, four brands
BrandAnswers naming brandVisibility rate (of 120 answers)Share of voice (of all mentions)
Your brand3025.0%18.9%
Competitor B6050.0%37.7%
Competitor C4537.5%28.3%
Competitor D2420.0%15.1%
Total mentions159n/a100%

Should you use a spreadsheet or a tool?

Start with a spreadsheet to learn what you are measuring. Move to a tool when running prompts by hand takes too long. Ahrefs Brand Radar tracks mentions, citations and AI Share of Voice across AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot 1.

Semrush Prompt Tracking checks a fixed set of prompts daily against one chosen platform, supporting ChatGPT, Google AI Mode and Gemini; it does not support Claude or Perplexity as of its 12 October 2026 article. It reports an AI visibility rate, defined as the share of tracked prompts where your brand appears, and its article does not mention a share of voice metric 7. It also warns that small groups are volatile and that a true 28-day comparison takes about eight weeks 7.

Whichever you choose, ask whether it reports share against named rivals or only your own visibility rate, how it samples, how often it runs and what it counts as a mention. Our tracking tool comparison goes further, and manual vs automated tracking helps you decide.

What can Google's own reports add?

Google's generative AI performance report in Search Console covers AI Overviews and AI Mode and reports impressions only, by page, country, date and device. Google says it rolled out to all websites worldwide as of 31 August 2026 8.

It does not show competitors, so it is not share of voice. Use it as a cross-check on your own Google exposure while your sheet covers the rest.

Which mistakes spoil the numbers?

Changing prompts mid-way is the top error, because it breaks every comparison. Semrush recommends keeping core prompts stable for at least a month 7. Add new prompts as a separate group instead.

Other errors: only testing prompts that name your brand, treating one lucky answer as a trend, and ignoring whether the answer is accurate. When you find a wrong claim, follow how to fix wrong information AI says about your brand.

When is share of voice not worth measuring?

Skip it when you are so new that no AI engine names you or your rivals. A zero tells you little. Spend the time on pages and mentions first, then measure.

Also skip it as your only metric. Pair it with leads from the sales process, as in how to measure AI SEO ROI.

Frequently asked questions

How many prompts do I need to measure AI share of voice?

Start with 30 to 50 across the buyer journey. Semrush notes that small groups are volatile: with six prompts in a group, one prompt change moves that group's rate by 17 points 7. Larger, stable sets give steadier trends.

Is AI share of voice the same as visibility score?

Not quite. Visibility is usually the share of tracked prompts where you appear. Share of voice compares your presence with competitors. Ahrefs defines it against rivals within a topic set 1, while Semrush defines visibility rate as share of prompts with your brand 7.

How often should I measure?

Monthly is enough for most small teams. Profound found that once a day was almost as precise as ten times a day 3. More frequent checks add cost without much extra signal. Keep the schedule and the prompts the same.

Can I track Claude or Perplexity?

Yes by hand. Among tools we checked, Ahrefs lists Perplexity and makes Claude available on its all-platforms option, where each Claude update uses more of your check allowance 1, while Semrush's Prompt Tracking did not support Claude or Perplexity as of 12 October 2026 7. Check each vendor's current list before buying.

Does a higher share of voice mean more leads?

Not necessarily. Share of voice measures presence, not buying. Pair it with self-reported attribution and sales data. We do not guarantee any link between the two for your business.

About the author

Rahil Jain

Rahil Jain is an AI SEO (answer engine and generative engine optimisation) consultant and co-founder of Chalk Labs.

LinkedIn · Method and credentials

Sources

  1. Ahrefs Brand Radar · AhrefsOpened October 20, 2026. Supports: Definitions from the FAQ: mentions (brand name in answer text), citations (AI links to your pages), AI Share of Voice (percentage of AI responses within a topic set that mention or cite your brand versus competitors).; platforms listed: AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot, with Claude on the all-platforms option.
  2. How to Measure AI Search Visibility When Attribution Falls Short · AhrefsOpened October 20, 2026. Supports: Track a sample of prompts and see who gets recommended; AI answers are inherently volatile; dated 6 October 2026.
  3. Is once a day enough? · ProfoundOpened October 20, 2026. Supports: 753 prompts, 7 US platforms, 1 to 14 June 2026: once a day within about 2 points of ten times a day; ten runs cut daily wobble by about 40% but platform drift remained; prompt portfolio matters more; vendor, internal test.
  4. Update: 38% of AI Overview Citations Pull From The Top 10 · AhrefsOpened October 20, 2026. Supports: AI Overviews are probabilistic and citations change with each query, so results are snapshots.
  5. How ChatGPT Does Query Fan-Out (15 Million Queries Studied) · AhrefsOpened October 20, 2026. Supports: Only 14.4% of 66.6 million free, logged-out ChatGPT responses (August and September 2026) triggered a web search.
  6. AI Visibility Is Lying to You: There's No One-Size-Fits-All AI Search KPI · Seer InteractiveOpened October 20, 2026. Supports: Published 31 August 2026; argues visibility counts positive and negative mentions alike; proposes AI Signal Rate, Answer Accuracy Rate and AI Influenced Conversion Rate; no formulas given.
  7. How to build an AI visibility dashboard with Semrush · SemrushOpened October 20, 2026. Supports: Published 12 October 2026: Prompt Tracking runs daily on one platform, supports ChatGPT, Google AI Mode and Gemini but not Claude or Perplexity; AI visibility rate definition; six-prompt groups move 17 points per prompt; keep core prompts stable a month; 28-day comparison takes about eight weeks.
  8. Generative AI performance report · Google Search Console HelpOpened October 20, 2026. Supports: Covers AI Overviews and AI Mode; impressions only; by page, country, date, device; rolled out to all websites as of 31 August 2026.
How this page was made. Built on Rahil Jain's AI SEO method and drafted with AI research assistance. Every statistic links to the page we read it on (opened October 20, 2026 or nearby). Published October 20, 2026; last updated October 20, 2026. Chalk Labs sells AI SEO services, so we have a commercial interest in this topic. Spotted an error? Tell us and we will fix it.