AI share of voice is the portion of AI answers that mention or cite your client compared with their competitors.
After reading this guide, you can measure it for any client. You will compare them against 3 to 5 competitors. You will also report the result honestly in your client reports.
What AI Share of Voice Means
AI share of voice is a simple ratio. Ask AI engines the same questions your client's buyers ask. Count how often each brand appears in the answers. Your client's count, divided by the total for all brands, is their share.
There are two things to count:
- Mentions: the answer names the brand.
- Citations: the answer links to or names the brand's page as a source.
Keep them separate. A mention shows awareness. A citation shows the engine trusted the page enough to point to it.
Be cited, not just ranked. A brand that gets named without a source has weaker footing than one the engine points to.
Why It Matters for Agency Reporting
Rankings show one list. AI answers show a short set of brands, often with no list at all. Clients ask whether they appear there. Your client reporting process needs a number that answers that question.
Here is what we measured for our own site. Bing Webmaster Tools AI Performance shows that Microsoft Copilot cited aiso.studio 13,854 times between 29 July and 27 October 2026, across 144 different questions. By month, that was 5,545 in August and 7,072 in September. October had 1,225 through 27 October.
Those are our own measured findings, and they cover those dates only. Your client's numbers will differ.
How the Measurement Works
The method has four parts. Each one removes a source of noise.
A Fixed Question Set
A question set is a fixed list of the questions your client's buyers ask AI engines. Write it once. Do not change it between checks. If you swap questions, you cannot compare one month to the next.
A Fixed Competitor List
Pick 3 to 5 competitors. Choose the brands your client actually loses deals to. Do not add a new rival halfway through. If you must, note the change in the report.
Dated Checks
Every check gets a date. AI answers change, so an undated number means little. Record the date, the engine, and the exact wording of the question.
Two Separate Counts
For each answer, log mentions and citations for every brand. Use one row per question and one column per brand.
How to Measure It: Step by Step
Follow these steps for each client.
- Write the question set. Use questions from sales calls, support emails, and search data. Aim for questions a real buyer would type.
- List 3 to 5 competitors. Confirm the list with the client.
- Run each question in the same engine. Use a fresh session. Do not log in with a personal history if you can avoid it.
- Log every brand that appears. Mark each one as a mention, a citation, or both.
- Add the date. Save the full answer text as proof.
- Calculate the share. Divide the client's count by the total count for all brands.
- Repeat on a schedule. Run the same set on the same cadence each time.
Common Mistakes
These errors ruin the comparison:
- Changing the questions between checks.
- Counting a mention and a citation as the same thing.
- Running one check and treating it as final.
- Leaving out the date.
- Picking competitors the client never meets in the market.
A Worked Example
This is an example only, not real data. Say you run 20 questions. A client appears in 6 answers. Competitor A appears in 8, and Competitor B in 6. The total is 20 brand appearances. The client's share is 6 of 20.
Now split that by citations. The client may be named often but cited rarely. That gap tells you where to fix content.
Why AI Answers Vary, and What to Do About It
ChatGPT's answers vary from person to person and day to day. No one can guarantee a citation. Report what you saw on a given date, not what will happen.
To see what ChatGPT says about a business today, use the free ChatGPT brand check at aiso.studio/chatgpt-brand-check. It covers ChatGPT only.
Because answers shift, run each question more than once. Look at the pattern across runs. A brand that appears in most runs of a question has a stronger position than one that appears once.
Reporting AI Share of Voice Honestly
Good client reports show the method before the number. State the question set, the competitors, the engine, and the dates. Then give the share.
The summary is the part every client reads first. Keep it plain. Say what you measured, what changed, and what you will fix next.
What to Put in the Report
- The question set and the date of each check.
- Mentions and citations, shown separately.
- The client's share against each competitor.
- A note that AI answers vary.
- The next actions.
What to Do When the Share Is Low
A low share points to content gaps. Engines cite sources they can trust, so start with the pages that answer your question set.
Then fix the content. A monitoring tool is a product that only reports what it sees. It does not change the pages. AISO Studio fixes the content. It scores pages from 0 to 100 on fact-check, AEO, SEO, readability, and engagement. Fact-check carries the most weight.
Frequently Asked Questions
How many questions should I track?
Start with a set you can run on a regular schedule. Use questions your client's buyers really ask. A short, stable list beats a long list that you abandon.
Should I count mentions or citations?
Count both, and report them separately. A mention means the brand was named. A citation means the answer pointed to the brand's page as a source.
What if the competitor list changes?
Note the change and the date in the report. Compare old and new lists with care. Where possible, keep the original rivals in the count.
What if the results differ from one run to the next?
That is normal, because AI answers vary. Run each question more than once and report the pattern. Never present one run as a fixed result.
Can anyone guarantee a client will be cited?
No. ChatGPT's answers vary from person to person and day to day. You can improve content quality and sources. You cannot control the answer.
Does AISO Studio track Gemini, Perplexity, or Copilot answers?
No. Our brand check covers ChatGPT only. The Copilot figures above come from Bing Webmaster Tools AI Performance for aiso.studio.
Key Takeaways
- AI share of voice is the client's share of mentions and citations across a fixed question set.
- Use 3 to 5 competitors and keep the list stable.
- Date every check and save the answer text.
- Count mentions and citations separately.
- AI answers vary, so report patterns and never promise citations.
- Explain the method in every client report, not just the number.
- Low share points to content you can fix.
Next Step
You now have a repeatable method for client reports. Run it on one client this week. Try the free content audit at aiso.studio/audit, then check what ChatGPT says today at aiso.studio/chatgpt-brand-check.