AI Search Share of Voice

Compare AI-search visibility for a target brand against competitors across every major LLM engine, then analyse the topic clusters each brand owns and where gaps exist.

Prerequisites

• SE Ranking MCP server connected.
• User provides: (a) target domain and its brand name, (b) list of competitor domains and brand names, (c) country (default: us), and (d) optionally, which engines to analyse (default: all supported: ai-overview, chatgpt, perplexity, gemini, ai-mode).

Process

1. Leaderboard snapshot DATA_getAiOverviewLeaderboard

   • Pull the AIO leaderboard for the target domain's category in the target country.
   • Capture mention counts and share percentages per engine, per domain.

2. Heatmap table

   • Build a table: rows = domains (target + competitors), columns = engines, cells = % share of voice.
   • Highlight the leader per engine and the worst performer.

3. Prompt sampling per domain DATA_getAiPromptsByBrand, DATA_getAiPromptsByTarget

   • For each domain (target and each competitor):
     • Pull 10 ChatGPT prompts where the domain appears as a source (link mention).
     • Pull 10 ChatGPT prompts where the brand is mentioned by name.
   • Save query text and the exact sources cited so the user can validate.

4. Topic clustering

   • Group prompts by theme (e.g., pricing, feature comparison, tutorials, alternatives, reviews).
   • For each brand, note which clusters it dominates and which it is absent from.

5. Gap and recommendation synthesis

   • Identify 3 to 5 topic clusters where the target underperforms competitors despite having relevant content.
   • Recommend specific actions: new content angles, structured data additions, partnerships with frequently-cited sources, comparison pages, FAQ/How-To schema.

Output format

Create a folder seo-ai-search-share-of-voice-{target-slug}-{YYYYMMDD}/ with:

seo-ai-search-share-of-voice-{target-slug}-{YYYYMMDD}/
├── 01-leaderboard.md         # raw leaderboard per engine
├── 02-heatmap.md             # visual heatmap table
├── 03-prompts-{domain}.md    # one file per domain with 20 sampled prompts
├── 04-topic-clusters.md      # cluster membership per brand
└── REPORT.md                 # executive summary

REPORT.md follows this shape:

# AI Search Share of Voice: {target brand} vs competitors

## Summary
- Target: {target} ({share}% across all engines)
- Leader: {leader brand} ({share}%)
- Target rank: {n} of {total}

## Heatmap

| Domain | AI Overview | ChatGPT | Perplexity | Gemini | AI Mode |
|---|---|---|---|---|---|
| {target} | {%} | {%} | {%} | {%} | {%} |
| {comp1} | ... | ... | ... | ... | ... |

## Who owns what

### {target brand}
Strong in: {cluster 1}, {cluster 2}
Absent from: {cluster 3}, {cluster 4}

### {competitor 1 brand}
...

## Topic cluster ownership

| Cluster | Leader | Share | Target position | Gap |
|---|---|---|---|---|
| Pricing | {brand} | {%} | {n} | {% behind} |
| Alternatives | {brand} | {%} | {n} | {% behind} |
| Tutorials | {brand} | {%} | {n} | {% behind} |

## Top 5 actions to close gaps
1. {action with target cluster}
2. ...

Tips

• Do not hallucinate citation counts. If the API returns zero prompts for a given domain/engine, report zero, do not estimate.
• For each competitor, validate the brand-name match in the prompt text. Sometimes "Wix" appears in a sentence about "wiktionary" or a person's name. Flag ambiguous matches in the raw-prompt file.
• base_domain scope is the default; do not narrow to subdomain unless the user asks.
• Respect Data API rate limit: 10 requests per second. With 5 domains and 2 prompt queries per engine per domain, pace the loop.
• The report is not a one-time artefact. Recommend the user re-run monthly and diff results to see ranking momentum.