Track AI Share of Voice Across LLMs Using Content Hub MCP
Retrieve AI share of voice data for a GEO cluster in Claude Desktop, then generate a prioritized content plan from it.
Share of voice is the percentage of all citations in a cluster, across every competing domain, that go to your brand. It differs from citation rate. Citation rate shows whether you are cited at all, and share of voice shows how much of the total citation volume you hold. With Claude Desktop connected to Content Hub MCP, you can retrieve this data for any GEO cluster and have Claude convert it into a prioritized content plan. If you have not connected Claude Desktop, follow the Content Hub MCP setup guide.
Before you start
- Create a GEO cluster and add prompts to it.
- Complete several GEO runs across the models you want to evaluate.
Fetch share of voice data for a GEO cluster
Request the AI share of voice data for the Developer marketing cluster across Claude, ChatGPT, Perplexity, and Gemini.
Claude locates the cluster and retrieves its prompt breakdown and competitor data. Share of voice is aggregated across the models that ran the prompts. It is not split per engine, so a per-model request can return a single combined figure if the cluster was not evaluated on every model.
Claude then returns the overall citation rate and share of voice for the cluster, and a table that ranks each competing domain by share of voice and mention count.
The response also summarizes your position before you request analysis. Your brand ranks in the top 3 on individual prompts and is strongest on agency-recommendation queries. Several prompts have a 0% citation rate, and a few prompts recently dropped in rank.
Analyze the data
Ask Claude to analyze the share of voice data in the conversation.
Claude opens with a one-line verdict. It then separates strong prompts from weak ones:
- Strong: bottom-of-funnel queries that ask for an agency recommendation.
- Weak or absent: educational and logistics queries, such as pricing or ROI.
The result is a funnel-stage gap rather than a broad content gap.
Claude then identifies the competitor with the most share of voice across prompts. This competitor is often a content or aggregator site rather than a direct competitor. Claude also lists the prompts that declined in rank and proposes a likely cause for each.
The analysis ends with three items:
- The prompts that improved, grouped to show the pattern to repeat.
- Zero-citation readings that are low-confidence rather than confirmed gaps.
- A priority order for action.
Request takeaways
Ask Claude for a condensed set of takeaways that you can scan or share. Claude returns a numbered list, and each item cites the supporting data:
- Your brand dominates decision-stage queries.
- Your brand is absent on informational queries. These provide the fastest wins.
- One recurring competitor is worth reverse-engineering.
- A few prompts are declining.
- Several prompts are improving and show what to repeat.
- A few zero-citation readings remain low-confidence.
The full workflow, from the share of voice data to a prioritized content plan, runs in one Claude Desktop conversation without manual cross-referencing across engines.
To review your current position against competitors, open app.infrasity.com and check the share of voice for your own cluster.
