/projects//ai_search/prompts/ ).
1. Defining the Conversational Funnel
Traditional search tracking counts keyword ranking positions (1-10). In conversational search, users receive synthesized answers containing 2 to 4 recommended options. Measuring whether your brand is option #1, an honorable mention, or omitted entirely is the new benchmark for search market share.
LIVE PLATFORM SCREEN & TELEMETRY
SEO Runway AI Search (GEO) Prompts & Visibility Index — Active Workspace
Below is the exact production screen with live parameters and dummy data illustrating the diagnostic findings and remediation buttons.
2. Setting Up Prompt Tracking in SEO Runway
Add up to 100 conversational prompts reflecting core buyer queries. SEO Runway automatically queries each model, parses the response text, and updates your composite GEO Visibility Score.
Target System: SEO Runway GEO Engine
Execution Frequency: Automated Background Queue (UTC Sync)
Tested across enterprise datasets with zero rate limiting or data loss.
3. What Share of Voice Means in Conversational Search
In classic search, share of voice measures how much of the visibility for a set of keywords your brand captures compared with competitors, usually weighted by ranking position and search volume. In conversational search, the units are different. There are no fixed positions or volumes, only answers. Conversational share of voice is the share of AI answers, across a defined set of prompts, that mention or cite your brand compared with competitors.
It answers a question leadership increasingly asks: "When our customers ask an AI assistant for a recommendation in our category, how often do we come up, and how often do our competitors?"
4. Building a Reliable Prompt Set
Share of voice is only as meaningful as the prompts behind it. Build the set carefully:
- Start from real questions. Use sales call notes, support tickets, site search logs, Search Console queries phrased as questions, and "People also ask" results.
- Cover the whole journey: problem-aware prompts ("how do I reduce no-shows at my clinic?"), solution prompts ("best appointment reminder software"), comparison prompts ("X vs Y"), and brand prompts ("is X worth it?").
- Add context variants for company size, industry, budget and location, since recommendations change with context.
- Weight by importance. Tag each prompt with a priority (for example 3 for high-intent commercial prompts, 1 for early research) so the score reflects business value, not just counts.
- Keep it stable. Change the set rarely and deliberately, so month-on-month comparisons stay valid. Add new prompts as a separate segment first.
Thirty to a hundred prompts per brand is typical. Smaller sets are noisy; much larger sets are expensive to run repeatedly without automation.
5. Calculating Share of Voice
For each prompt, run it several times in each engine you track and record which brands are mentioned and which are cited. Then calculate:
- Mention rate: the percentage of runs in which your brand is mentioned.
- Share of voice: your mentions divided by total mentions of all tracked brands (you plus a defined competitor set) across all runs, optionally weighted by prompt priority.
- Citation share: the same calculation using citations (links to your domain) instead of mentions.
- Position-weighted share: where answers list brands, give more weight to earlier positions (for example 1.0 for first, 0.7 for second, 0.5 for third).
Report each engine separately as well as combined. Google AI Overviews, ChatGPT, Gemini and Perplexity often produce very different shares for the same brand.
6. Dealing with Variability
AI answers are non-deterministic: the same prompt can produce different brand lists from one run to the next, and answers shift as models and retrieval systems update. To get stable numbers:
- Run each prompt multiple times per measurement period (three to five runs is common) and average the results.
- Use fresh sessions without conversation history, with consistent location and language settings.
- Track rolling averages over four weeks rather than reacting to single weekly readings.
- Note known model or product updates on your trend charts, so sudden shifts can be explained.
- Look at confidence, not just averages: a share of voice that swings between 10% and 40% week to week is far less reliable than one moving between 22% and 26%.
7. Choosing the Competitor Set
Share of voice depends on who you compare against. Define a fixed set of five to ten competitors per category, including direct competitors you meet in sales, category leaders that AI answers mention frequently (even if you don't see them as direct rivals), and emerging challengers. Review the set quarterly. If AI answers repeatedly mention a brand you aren't tracking, add it, because your real share of voice is lower than your report suggests.
8. Turning Share of Voice into Action
- Find the gaps: prompt segments where competitors dominate and you're absent.
- Check the sources: for those prompts, which pages and sites are cited? Review sites, forums, comparison articles and documentation are common.
- Fill content gaps: create or improve pages that directly answer those prompts with clear, factual, current information.
- Build third-party presence: earn inclusion and accurate descriptions on the sources engines cite for those prompts.
- Re-measure the affected segment monthly to see whether actions moved the share.
9. Reporting Share of Voice
A clear monthly report includes:
- Headline share of voice across all engines, with the trend over the last six months.
- Share by engine (AI Overviews, ChatGPT, Gemini, Perplexity), since strategies differ by engine.
- Share by prompt segment (research, comparison, recommendation, brand) to show where you win and lose.
- Competitor comparison, showing which brands gained or lost share.
- Citation share alongside mention share, because citations send traffic while mentions build awareness.
- Actions and next steps tied to specific gaps.
Put conversational share of voice next to classic search share of voice for the same topics. Leadership can then see the whole search picture in one place, and spot cases where you rank well in classic results but are rarely recommended in AI answers, or the reverse.
10. Connecting Share of Voice to Business Outcomes
Share of voice is a leading indicator, not revenue. Connect it to outcomes where you can: referral sessions and conversions from AI platforms in analytics, branded search volume trends (AI recommendations often lead users to search for the brand afterwards), and "how did you hear about us?" responses that mention AI assistants. None of these is perfect on its own, but together they show whether rising conversational visibility translates into demand.
11. Manual Tracking vs Automated Tools
For a small prompt set, manual tracking in a spreadsheet works: run each prompt, paste the answer, and tag brands mentioned and cited. It quickly becomes impractical, though. Fifty prompts, four engines and five runs each is a thousand answers per measurement cycle. Automated AI visibility tools run prompts on a schedule, detect mentions and citations, identify competitors named, and chart share over time. When choosing one, check which engines it covers, whether it separates mentions from citations, how it handles repeat runs and variability, and whether it lets you use your own prompts and competitor set.
12. Common Mistakes
- Too few prompts, so a single answer swings the whole score.
- Changing the prompt set every month, which makes trends meaningless.
- Counting mentions only and ignoring whether you are cited or described positively.
- Leaving out the competitors AI actually recommends, which inflates your share.
- Combining engines into one number and missing engine-specific problems.
- Reacting to weekly noise instead of rolling trends.
13. Share of Voice Checklist
- 30–100 prompts from real customer questions, across every journey stage.
- Prompts weighted by business priority and kept stable.
- A fixed competitor set reviewed quarterly.
- Multiple runs per prompt per engine, in fresh sessions.
- Mention share, citation share and position-weighted share calculated.
- Results reported by engine and by prompt segment, with rolling trends.
- Gaps traced to cited sources and turned into actions.
- Share of voice connected to AI referral traffic, branded search and conversions.
14. A Worked Example
A regional accounting firm tracks 40 prompts such as "best accountant for small businesses in Bristol" and "do I need an accountant for a limited company?", plus five local competitors. Across four engines and three runs per prompt, there are 480 answers in a month. The firm is mentioned in 58 of them, while its largest competitor appears in 141. Mentions of all six tracked firms total 400.
- Mention rate: 58 ÷ 480 = 12%.
- Share of voice: 58 ÷ 400 = 14.5%, against the leading competitor's 35%.
- Segment view: the firm does well on informational prompts (where its guides are cited) but appears in only 4% of "best accountant in Bristol" answers, where review platforms and local directories dominate citations.
The action plan follows directly: improve and expand reviews on the platforms cited for local recommendations, correct inconsistent listings in local directories, and add a clear service-area page for Bristol with fees and specialisms. Three months later, the firm's share on local recommendation prompts can be compared against this baseline to judge whether the work paid off.
15. Making It a Habit
Conversational share of voice becomes valuable when it is measured the same way every month and reviewed alongside your other marketing metrics. Assign an owner, fix the prompt and competitor sets, automate the runs where possible, and spend the review meeting on the gaps and the actions, not on individual answers. Over a few quarters, the trend line tells a clearer story about your brand's place in AI-driven discovery than any single test ever could.
16. Share of Voice for Agencies Managing Many Brands
Agencies tracking conversational share of voice for several clients should standardise the method so results are comparable and efficient to produce. Use one prompt-building template for every client (journey stages, context variants and priority weights), one scoring definition for mentions, citations and sentiment, and one report layout. Keep each client's prompt and competitor sets in a separate project so data never mixes. Present conversational share of voice next to the client's classic rankings and organic traffic in the same monthly report, so they see one search story rather than separate AI and SEO updates. Clients quickly understand a simple line: "Across 50 questions your customers ask AI assistants, you were recommended in 18% of answers this month, up from 12% last quarter."
Want the complete procedural runbook?
Visit our comprehensive step-by-step Knowledge Base for interactive screen maps, configuration parameters, and adjacent module workflows.
Open Knowledge Base Hub →

Bookmarked. We run this exact setup in SEO Runway across our agency client retainers. As a VP Marketing, CloudScale, this provides exactly the architectural depth we look for.