These six prompts measure competitive AI share of voice (SOV) against named rivals. In AI search, answer share of voice is your share of all tracked-brand mentions. You copy each prompt, fill the [placeholders], and run it.
The recipes cover ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews. Four prompts calculate and compare SOV across those engines. The last two turn gaps into actions and write your monthly readout.
Every recipe uses one formula. Competitive SOV = your brand mentions ÷ all tracked-brand mentions in the same answer set × 100. Keep that formula fixed across engines and months. A new formula resets every trend line.
Each recipe ends with a short list of checks. Run those checks before you share a number. Bad inputs produce confident but wrong SOV figures.
The scope stays tight on purpose. You get measurement and planning prompts for your own search program. For a full competitive audit, use our competitor analysis prompts.
You can run every prompt by hand in any chat tool. AirOps tracks share of voice against named competitors across ChatGPT, Gemini, Perplexity, and Claude.
- Fix your competitor set and prompt panel before you count anything.
- Run each prompt 3+ times per engine, because answers shift between runs.
- Calculate SOV per engine, then aggregate from raw mention counts.
- Split mentions from citations to see why a rival wins.
- Turn each gap into one action, and report the trend monthly.
How do I check share of voice in AI-powered search?
Fix a prompt panel and competitor set, run each prompt 3+ times per engine, count mentions, and divide. Those four steps show how to measure AI share of voice.
- Freeze one non-branded prompt panel and one named competitor set.
- Run every prompt 3+ times in each AI engine.
- Count mentions of each tracked brand in every answer.
- Divide your mentions by all tracked-brand mentions, then multiply by 100.
Repeat runs matter because answers drift. AirOps brand visibility research found only 20% of brands stay present across five consecutive runs.
One run gives you a snapshot. Three or more runs give you a share you can trust. Average the counts across runs before you divide.
Count a mention when the answer text names the brand. Source lists and links go to the citation split in Prompt 4.
Log each answer with its prompt, engine, run, and date. Save the full raw answer text, never a short summary. That single log feeds every prompt in this library.
Separate branded prompts from non-branded prompts. Branded prompts name you, so they inflate your share. Competitive SOV uses non-branded prompts only.
Calculate SOV per engine first. Each engine draws on different sources. Build the aggregate from raw counts, never from averaged percentages.
This page assumes you know the metric already. The share of voice guide covers SOV across every marketing channel.
Six prompts to measure competitive AI share of voice
Run the six prompts in order, because each one feeds the next. These AI SOV prompt templates work in any large language model (LLM) chat. Paste your exported AI answers wherever a prompt asks for them.
- Prompt 1 builds the competitor set and prompt panel.
- Prompt 2 calculates competitive SOV from your answer export.
- Prompt 3 compares SOV across five AI engines.
- Prompt 4 splits mentions from citations.
- Prompt 5 turns SOV gaps into ranked actions.
- Prompt 6 writes the monthly readout.
Prompt 1 to build your competitor set and prompt panel
Run this prompt first. It builds a prompt set for AI visibility that stays fixed. The table sets the rules for each field.
| Field | Rule | Example |
|---|
| Competitor set | Named brands only, fixed for the quarter | [competitor A], [competitor B] |
| Prompt panel | Non-branded buyer questions, frozen after approval | Which [category] tool fits [use case]? |
| Funnel stage | One tag per prompt | Problem, comparison, or recommendation |
| Branded flag | Mark prompts that name any tracked brand | Yes or no |
| Engines | Same five engines every run | ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews |
You are a search analyst. My brand is [your brand] in [category]. My buyer is [persona] at [company type]. List [number] competitors this buyer compares with [your brand].
Then write [number] non-branded prompts this buyer asks AI assistants. Mix problem, comparison, and recommendation prompts. Tag each prompt with its funnel stage and topic. Return a table with prompt, stage, topic, and branded flag.
Review the panel before you lock it. Look for these four signals.
- Every competitor is a brand your buyers name.
- No prompt names your brand or a rival.
- Each funnel stage has prompts in the panel.
- The saved panel carries a version date.
Prompt 2 to calculate competitive SOV
This prompt applies the AI share of voice formula to your answer export. It counts mentions per brand, then divides by all tracked-brand mentions. The core answer engine metrics add citation rate and sentiment.
| Input | Definition | Example |
|---|
| Brand mention | The answer text names the brand | An answer names [your brand] |
| Answer set | All answers from one panel, engine, and period | [engine], [month] |
| Your mentions | Answers in the set that name you | 18 (illustrative) |
| All tracked-brand mentions | Mentions of every tracked brand in the same set | 72 (illustrative) |
| Competitive SOV | Your mentions ÷ all tracked-brand mentions × 100 | 18 ÷ 72 × 100 = 25% |
The example numbers are illustrative only. You earn 18 mentions, and all tracked brands earn 72. Your competitive SOV is 25%.
Act as a measurement analyst. Below are AI answers for my prompt panel. Tracked brands are [your brand], [competitor A], [competitor B], and [competitor C]. Exclude answers from branded prompts.
Count each brand once per answer that names it. Calculate SOV as brand mentions ÷ all tracked-brand mentions × 100. Return counts and SOV per brand for these answers: [paste answers]
Audit the math before you trust it. Confirm each item below.
- Each brand counts once per answer, even when named twice.
- Branded prompts stay out of the calculation.
- Brand aliases and product names map to one brand.
- Shares across tracked brands sum to 100%.
Prompt 3 to compare SOV across AI engines
This prompt shows your share of voice across ChatGPT, Gemini, Perplexity, and Claude. It adds Google AI Overviews as a fifth engine. Google says the Gemini app passed one billion monthly users in August 2026.
Engine audiences differ in size, and each engine cites different sources. So competitor share of voice in AI answers shifts by engine. One blended number hides where each rival wins.
Act as a search analyst. Here are SOV counts per engine for [period]. Engines are ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews.
Build one table with engine, my mentions, [competitor A], [competitor B], SOV, and runs. Flag engines where I trail the leader. Build the all-engines row from raw counts. Data: [paste counts]
| Engine | Your mentions | Competitor A | Competitor B | Your SOV | Runs |
|---|
| ChatGPT | [count] | [count] | [count] | [%] | [3+] |
| Gemini | [count] | [count] | [count] | [%] | [3+] |
| Perplexity | [count] | [count] | [count] | [%] | [3+] |
| Claude | [count] | [count] | [count] | [%] | [3+] |
| Google AI Overviews | [count] | [count] | [count] | [%] | [3+] |
| All engines | [sum] | [sum] | [sum] | [%] | [total] |
Read the output engine by engine. Check these points first.
- Every engine ran the same panel and run count.
- The all-engines row sums raw counts.
- Gaps above [threshold] points carry a flag.
- Prompts with no AI Overview stay out of that row.
Prompt 4 to split mentions from citations
A mention names your brand in the answer. A citation links to your page as a source. This mentions and citations breakdown explains both signals, and our LLM citation audit shows which pages AI cites.
Mention rate vs citation rate shows why a rival wins. The prompt sorts every answer into one of four quadrants.
Act as a citation analyst. For each answer below, mark whether it names [your brand]. Also mark whether it cites a [your domain] URL.
Do the same for [competitor A] and [competitor B]. Sort each brand's answers into four quadrants. Report mention rate, citation rate, and quadrant counts per brand. Answers: [paste answers]
| Quadrant | What it means | Next step |
|---|
| Mentioned and cited | AI names you and links your page | Protect and refresh the cited page |
| Mentioned only | AI knows you but sources other sites | Earn a citation on the topic |
| Cited only | AI uses your page but names a rival | Name your brand clearly on the page |
| Absent | AI skips you on the prompt | Build owned and third-party coverage |
Grow the mentioned-and-cited quadrant first. AirOps citation and mention research found brands earning both were 40% more likely to resurface across runs. That lift compares against brands earning citations alone. To plan which brands you appear beside, use our brand co-mention prompts.
Run Prompt 4 on the same answer set as Prompt 2. Mixed sets break the quadrant math.
Validate the split before you act on it. Confirm these four rules.
- Citations count only links to your domain.
- Mention rate and citation rate stay separate metrics.
- Each answer lands in exactly one quadrant.
- Cited-only answers list the page AI used.
Gap-to-action prompt for growing AI share of voice
Prompt 5 turns an SOV gap into a ranked action plan. It reads your Prompt 3 and Prompt 4 outputs. Then it matches each gap pattern to its usual cause.
Start with the engines where you trail most. Sort next by quadrant, because each quadrant needs a different fix. One action per gap keeps the plan short.
Act as a search strategist. Here are my SOV results by engine and quadrant for [period]. Competitors are [competitor A] and [competitor B].
For each gap, name the pattern, the cause, and one action. Rank actions by gap size, then effort. Tie each action to one prompt cluster and one page or source. Results: [paste outputs]
Pressure-test the plan before you assign work. Check these three points.
- Every action names one page or source.
- Actions rank by gap size first.
- Each gap maps to one cause.
Fewer clicks raise the stakes of each gap. Pew Research Center measured Google clicks in a 2025 click behavior study. Searchers who saw an AI summary clicked a traditional result in 8% of visits. Without a summary, they clicked one in 15% of visits.
| Gap pattern | Usual cause | Action |
|---|
| Low SOV on one engine only | That engine cites sources that skip you | Earn coverage on the sources that engine cites |
| Mentioned only, rarely cited | No owned page answers the prompt cleanly | Publish or restructure a page for the prompt |
| Cited only, rarely mentioned | Your page names your brand weakly | Add brand and product names near each answer |
| SOV drop month over month | Cited pages went stale | Refresh the page and its data |
| Absent across engines | Thin owned and third-party coverage | Build third-party mentions first |
Pick one row per engine gap. Assign an owner and a due date. Recheck SOV in the next monthly run.
Offsite sources shape your share. The AirOps 2026 State of AI Search report found 85% of brand mentions originate from third-party pages. The same report found pages not updated quarterly are 3× more likely to lose citations. Our offsite signals report and Reddit citation prompts help you plan third-party actions alongside page refreshes.
Content changes move generative visibility. Aggarwal and colleagues tested generative engine optimization (GEO) in a 2024 peer-reviewed study. GEO methods boosted visibility by up to 40% in generative engine responses.
Monthly AI share of voice readout prompt
Prompt 6 writes your monthly AI share of voice report. The output fits one page for your SEO stakeholders. It reuses the outputs from Prompts 2 through 5.
Act as an SEO analyst writing a monthly readout. Use my SOV, quadrant, and action outputs for [month] and [prior month].
Write five short sections on SOV trend, engine changes, competitor moves, quadrant shifts, and next actions. Keep sentences under 15 words. Flag any change in panel, engines, or competitors. Data: [paste outputs]
Scan the draft before it goes out. Confirm these three points.
- The trend compares the same panel month over month.
- Each competitor move cites the prompts behind it.
- Next actions match your Prompt 5 plan.
Add AI SOV benchmarks for context. The answer engine metrics guide from Prompt 2 sets three bands. Calibrate them to your competitor set.
- 0 to 5% marks an early program.
- 5 to 20% marks a competitive position.
- 20%+ marks a category leader.
SOV is one line in a wider report. This visibility measurement guide covers the rest of the metric set.
Rules that keep competitive AI share of voice comparable
Comparable numbers need fixed inputs. One changed input breaks your trend line. Hold these five rules every month.
- Freeze the prompt panel for the quarter and version every change.
- Keep branded prompts out of competitive SOV.
- Run each prompt 3+ times per engine per period.
- Build aggregates from raw counts, never averaged percentages.
- Keep the same competitor set across engines and months.
Log every exception in the readout. A new rival or a new engine starts a new baseline.
Key takeaways
- One fixed formula keeps your SOV trend honest.
- Per-engine numbers show where each rival wins.
- The mention and citation split points to the right fix.
- A monthly readout turns SOV into a standing work queue.
Frequently asked questions
How do I monitor competitor mentions in AI search results?
Run a fixed non-branded prompt panel 3+ times per engine each month, and log every brand each answer names. Prompt 2 counts those mentions per competitor, and Prompt 3 shows which engine each rival wins. To compare services that report this for you, see our enterprise AI visibility services prompts.
How many times should I run each prompt?
Run each prompt at least 3 times per engine per period. Answers shift between runs, so single runs mislead.
How often should I measure competitive AI share of voice?
Measure monthly for trend reporting. Rerun Prompt 5 after any major content change or competitor launch.
AirOps for competitive AI share of voice measurement
AirOps is an enterprise AEO platform and expert partner for marketing leaders, connecting AI search intelligence, coordinated action across channels, and measurement tied to growth.
AirOps Analytics sits inside the intelligence and measurement platform. It handles three jobs from this page.
- It tracks SOV by AI provider and named competitor.
- It separates mentions from citations in every answer set.
- It connects SOV gaps to content refresh workflows.
Your formula stays the same in AirOps Analytics. So do your panel rules and competitor set. AirOps Analytics adds a fixed run schedule and a trend history. Each readout then starts from tracked data.
AirOps experts help you rank the gaps worth closing first. They tie each action to a prompt cluster and a page. They also cover the third-party sources behind most brand mentions. Your monthly readout then shows whether SOV moved.
Bring your competitor set and prompt panel to the call. An AirOps expert walks through your SOV by engine with you. You see your largest gaps by engine and competitor.
Map your competitive AI share of voice when you book a call
Run the AEO strategy that wins AI discovery
AirOps works with you to set and execute the right AI search strategy to drive growth for your brand.