October 12, 2026
A practical method for measuring AI share of voice: pick tracked competitors, freeze a prompt set, run clean repeat samples, and combine ChatGPT, Gemini, and Perplexity scores without hiding the gaps.
Measure AI share of voice one platform at a time: your mentions divided by the mentions of every brand you track. Combine platforms only after.
Freeze the competitor list and prompt set, repeat every prompt 3 to 5 times in clean sessions, and log mentions apart from citations.
Read the trend over 60 to 90 days, because small monthly moves are mostly noise.
Ask ChatGPT the same buying question twice and you will rarely get the same shortlist back. Buyers act on whichever list they see, so one screenshot says almost nothing about your AI visibility.
Share of voice turns shifting lists into one comparable number, but only if you count the same way every time. A loose denominator, a single run per prompt or careless averaging gives you a figure that looks precise and says something false.
AI share of voice is the percentage of brand mentions in AI answers that belong to you, out of all mentions of the brands you track. If your brand takes 60 of the 300 mentions across tracked brands, your share is 20%. It is brand monitoring for AI answers, reduced to one number.
Traditional share of voice compared your slice of impressions or ad spend with your competitors'. An AI answer names a handful of brands and stops. There is no page two, so every mention a competitor gains is one you do not get. Our guide to AI share of voice marketing metrics covers the marketing background.
Three numbers get confused under this label:
Mention share: your mentions ÷ all mentions of tracked brands. This is the share of voice number.
Citation share: your domain's links ÷ all links to tracked domains. It shows who gets credited, not who gets named. Some tools divide by all cited links, which reads lower.
Mention rate: the percentage of answers that name you at all. It ignores competitors: named in 90% of answers beside four equally visible competitors, you still hold only 20%.
Make mention share your primary metric and write the denominator beside it: tracked brands only. Keep mention rate as a stability check. In the SparkToro and Gumshoe.ai study, how often a brand appeared held steady between runs, while its list position did not.
Measure each platform on its own before combining anything. Citation analyses from 2026 put the overlap between the domains ChatGPT and Perplexity cite at only about 11%, and a blend hides that: a 35% share on one platform and 10% on another average into a figure that describes neither.
Choose 3 to 6 direct competitors from your sales calls and lost deals, map each one's aliases to its parent brand, and fix your counting rules before the first run. Buyers compare you with these brands, so AI answers will too. With fewer than three, one competitor's good month swings your share alone; with more than six, the sheet is hard to maintain.
For each company, list the name as written, the spaced version, the bare root name and the domain, all mapped to one parent brand. Add a product name only if it is unmistakable. Generic ones such as "Pro" or "Earn" match unrelated text and create false positives, so Verseodin leaves product names out of its brand tokens entirely.
Three counting rules keep the data clean:
One mention per brand per answer: a brand named five times in one response counts once.
Mentions and citations stay separate: a brand named in the text is a mention, and a link to its domain is a citation. Log them in different columns.
No edits mid cycle: change a rule halfway and earlier numbers stop being comparable.
The competitor list moves your score. Say you and four competitors draw 300 mentions in ChatGPT and 60 are yours: 20.0%. Add a fifth competitor with 40 mentions and you fall to 17.6% (60 ÷ 340), though ChatGPT treats you exactly as before. Leave out your biggest competitor, who earned 85, and you jump to 27.9% (60 ÷ 215). Adding a competitor lowers everyone's share; missing one inflates yours.
Log any unexpected brand in an "other brands" column rather than adding it on the spot. Once a quarter, promote the names that keep appearing into the core set and mark the reset date on your trend chart, so nobody reads the shift as a decline.
How to measure share of voice in AI search starts with prompts that read like real buyer questions, in enough volume to hold steady. Use 30 to 50 prompts for a manual audit and 100 to 200 for ongoing tracking. Verseodin's standard setup runs about 100 unbranded prompts per Query Universe, with no brand or competitor names inside.
Source them from where buyers already talk:
Sales calls: questions prospects ask before naming a vendor.
Support tickets: problems customers describe in their own words.
Search Console: queries phrased as questions.
Reddit: threads asking for recommendations.
Review sites: complaints and comparisons about tools in your category.
Our guide to building and maintaining a buyer prompt library shows how to cluster, update and retire prompts over time.
Mix four prompt types, weighted toward shortlist and comparison. For a 50 prompt set:
Discovery (10): a problem, no category yet. "How do I know if AI assistants recommend my company?"
Use case (20): a situation that calls for a shortlist. "Which tools help a SaaS team of five track AI visibility on a small budget?"
Comparison (13): options weighed against criteria. "How do AI visibility tools differ on prompt volume and reporting?"
Purchase intent (7): a buyer close to deciding. "What should I check before paying for an AI visibility platform?"
Leave out any prompt that names your own brand; it guarantees a mention and inflates the score. Prompts naming a competitor skew the count differently: "alternatives to Competitor A" removes A from the answer, while "Competitor A vs Competitor B" guarantees both appear. Track them as a separate subset.
Three mistakes skew results most:
Short keywords instead of questions: "CRM software" is not how people ask an assistant.
Only "best X" list prompts: long lists swell the denominator, so the score measures list inclusion, not recommendation.
Changing prompts between cycles: you cannot tell a real move from a rewrite. Freeze the set between quarterly resets and log every edit.
Run the same prompt set on every platform, in clean sessions with no chat history, and run each prompt 3 to 5 times per platform. Answers change between runs: in a SparkToro and Gumshoe.ai study, 600 volunteers ran 12 prompts across ChatGPT, Claude and Google AI, producing 2,961 responses. The odds of the same brand list appearing twice were under 1 in 100, and about 1 in 1,000 for the same list in the same order.
A clean session means a logged out browser, private window or stateless API call, with personalization off and the same web search setting, location and language on every run.
Three to five runs is a practical floor, not a statistical guarantee. Verseodin runs each prompt daily on ChatGPT, Gemini and Perplexity, about 30 runs per prompt per month, so a monthly roll up is far steadier than a hand audit.
Recommended Tool
Don't guess whether ChatGPT, Claude, Gemini, or Perplexity can access your content. Analyze your site in seconds.
Check AI Visibility Try Prompt Finder
No signup required • Instant results
Give your tracking sheet eight columns:
Prompt: exact text and ID.
Platform: ChatGPT, Gemini or Perplexity.
Run: 1 to 5.
Brands named: every tracked brand, plus unexpected ones for the "other brands" column.
Position: where each brand appears in the answer.
Sentiment: positive, neutral or negative framing.
Cited domains: every domain linked in the answer.
Date: the run date.
Position is the noisiest column. SparkToro described list position as effectively meaningless, so read it as an average across runs, never as a single result.
A note on Google AI Overviews: they lean on links more than names. A wider BrightEdge dataset showed more than twice as many citations as brand mentions in AI Overviews, and in ecommerce queries brands were named in just 6.2% of AI Overviews against 99.3% of ChatGPT responses. Log cited domains there and report citation share.
How to measure share of voice across multiple AI platforms simultaneously comes down to one rule: calculate each platform first, then combine the results with a method you state up front. The formula never changes:
AI share of voice = your brand mentions ÷ total mentions across tracked brands × 100
The illustrative numbers below run through this section: you and four competitors, 50 prompts, 3 runs each, so 150 answers per platform.
Brand
ChatGPT
Perplexity
Gemini
You
60 (20.0%)
70 (35.0%)
10 (10.0%)
Competitor A
85 (28.3%)
45 (22.5%)
35 (35.0%)
Competitor B
70 (23.3%)
40 (20.0%)
25 (25.0%)
Competitor C
55 (18.3%)
30 (15.0%)
20 (20.0%)
Competitor D
30 (10.0%)
15 (7.5%)
10 (10.0%)
Total
300
200
100
Three ways to combine the platforms give three different answers:
Pooled: add every mention across platforms, then divide: (60 + 70 + 10) ÷ (300 + 200 + 100) = 23.3%. This weights each platform by how many brands it names, not by how many of your buyers use it.
Audience weighted: multiply each platform's share by the portion of your buyers who use it. At 60% ChatGPT, 30% Perplexity and 10% Gemini: (20.0 × 0.6) + (35.0 × 0.3) + (10.0 × 0.1) = 23.5%.
Simple average: (20.0 + 35.0 + 10.0) ÷ 3 = 21.7%.
The simple average misleads because it gives Gemini the same vote as ChatGPT, whether or not your buyers use it. At a 10/10/80 audience split the weighted score falls to 13.5%, yet the average still says 21.7%. Take the weights from GA4 referrals (chatgpt.com, perplexity.ai, gemini.google.com) and a "which AI tool did you use?" question on your demo form.
Calculate citation share from the same data (your domain's citations ÷ all citations to tracked domains) and report it beside mention share, never blended into it.
Platform
Your citations
All tracked citations
Citation share
Mention share
ChatGPT
6
50
12.0%
20.0%
Perplexity
38
100
38.0%
35.0%
Gemini
2
25
8.0%
10.0%
Pooled
46
175
26.3%
23.3%
Audience weighted (60/30/10)
19.4%
23.5%
ChatGPT gives you a 20.0% mention share but only a 12.0% citation share, which fits BrightEdge's finding that it mentions brands more than three times as often as it cites them. The pooled citation share of 26.3% leans on Perplexity, which supplies 100 of the 175 citations. Verseodin's GEO Audit divides by all cited links rather than tracked domains, so its citation share reads lower.
Scores differ by platform because each retrieves from the web differently, links at its own rate and prefers different sources. ChatGPT leans on reference and editorial sources such as Wikipedia, while Perplexity cites two to three times more sources per answer and favors community discussion and fresher pages. Our breakdown of how each platform sources and cites content goes further.
Competitor Scan Live preview
See which brands AI recommends ahead of you, and why.
Look for three patterns in your numbers:
High mentions, low citations: buyers hear your name without being sent to your site, as with ChatGPT above. Check which pages and third party sources the model could cite instead.
Strong on one platform, absent on another: 35.0% on Perplexity against 10.0% on Gemini usually points to a source gap on the weaker platform, not a brand problem everywhere.
Named often, low in the list: confirm it across runs before acting, because position is noisy.
Run the full prompt set monthly and a priority subset of 10 to 15 revenue prompts weekly, and judge trends over 60 to 90 days. Treat small month to month moves as noise. At a 30% appearance rate, pinning down how often one prompt names you to within ±10 points at 95% confidence takes about 81 runs, and comparing two months widens that margin to about 14 points. At hand audit volumes, a 2 point gain is not a result.
Record the prompt set version, run count, platforms, dates and competitor list with every score so quarters stay comparable.
Then turn the gaps into work:
Blindspots: prompts where competitors appear and you get neither a mention nor a citation. Fix the highest intent ones first.
Third party sources: open the cited domains behind competitor mentions; the reviews and roundups they appear in are where you need a presence.
Measure again: rerun the same set after 60 to 90 days against your frozen baseline.
Run the same unbranded prompts 3 to 5 times in ChatGPT, log every brand named, and calculate mention share. For AI Overviews, check priority prompts by hand and log the cited domains, since that surface links more than it names. Report the two separately.
There is no fixed number, because the count of tracked brands sets the baseline. With five brands, an even split is 20%, which means parity, not strength. Judge yourself against your closest competitor and your own 60 to 90 day trend.
Only if the denominator, prompt set, platforms and run counts match. A tool that divides by all cited links, or tracks different competitors, produces a different figure for the same brand. Treat outside numbers as direction only.
A spreadsheet works for a first audit. Fifty prompts across three platforms with three runs each is 450 answers, roughly 7 to 8 hours of copying at a minute each. Beyond that, automation pays for itself.
Keep the headline score unweighted: one mention is one mention. Log position and sentiment beside it as diagnostics, and average position across runs, since it shifts between identical ones.
TL;DR
What Is AI Share of Voice and Which Version Should You Measure?
How to Choose Competitors and Set Counting Rules
How to Select Prompts for Measuring Share of Voice in AI Search
How to Run Prompts and Track Mentions Across ChatGPT, Gemini, and Perplexity
How to Calculate AI Share of Voice Across Multiple AI Platforms Simultaneously
Why Scores Differ by Platform and How to Track Them Over Time
Frequently Asked Questions
Summarize with ChatGPT Summarize with Claude Summarize with Perplexity
S
Satvik Mishra
Co Founder of Verseodin
Satvik Mishra is the Co Founder of Verseodin, an AI visibility platform that tracks brand citations across ChatGPT, Gemini, Claude, and Perplexity. He writes about generative engine optimization strategy and what actually works for brands trying to earn visibility in AI powered search.
Newsletter
Get the latest AI Visibility insights, GEO research, product updates, and SEO strategies delivered straight to your inbox.
Continue improving your AI visibility.
Discover what users are asking AI about your industry and uncover prompt opportunities.
Open Prompt Finder
Analyze whether AI search engines can discover and cite your website.
Run Free Report
Tutorials 12 min read
Ten criteria for judging an AEO insights platform, from prompt tracking and brand mention analysis to bot data, citation sources, share of voice, and platform coverage, plus a way to choose.
Read article
Tutorials 11 min read
A practical guide to building and managing an AI search visibility strategy: audit the foundation, optimize high intent pages, track four KPI groups, analyze citation gaps, and run an ongoing optimization cycle.
Read article
Tutorials 12 min read
A mention count cannot explain ChatGPT visibility. Here is what an AEO tool should record on every run: the prompt, the full answer, competitors, products, citations, and history, and how to read them together.
Read article
Want more AI SEO insights?
Monthly research on AI search, GEO and citation trends. No noise.
[ partner program ]
Join our partner network, whether you're an agency, consultant, or reseller. Leave your email and we'll reach out with details.
Revenue share
Earn a cut of every referral you bring on, renewing month over month.
Co-marketing
Joint webinars, case studies, and content with the Verseodin team.
Priority support
Direct line to our team plus early access to new features.
How to Measure AI Share of Voice Across ChatGPT, Gemini, and Perplexity | VerseOdin