Most AEO pitches start the same way. The agency runs a few prompts about the client's brand, screenshots the gaps, and walks in saying "you are missing here, here, and here."
It works. It also frames the conversation around the client's individual visibility, which is the smaller question. The bigger question is: what does the category look like in AI search? Who is cited, who is not, where are the gaps that nobody owns? The brand-by-brand audit cannot answer that. The Visibility Index can.
This post explains what it is, why it changes the discovery call, and the exact way I run it.
What is the Visibility Index
The Visibility Index is a brand-agnostic view of an industry. You give it a name (e.g. "Project management for engineering teams"), an industry, and a set of topic clusters with prompts. It runs those prompts across ChatGPT, Claude, Perplexity, and Google AI Overviews and stores every entity cited as a competitor. There is no "your brand" filter. Nothing is privileged.
What you get back is three views the brand-level audit cannot give you:
- Share of voice: ranked entity list by citation frequency, with a concentration score (top-1 share, plus an HHI-style index showing how dominated the category is by one or two players).
- Sources: the domains the LLMs lean on most when answering prompts in this category.
- Whitespace: prompts and topics where no entity is consistently cited, the gaps where authority is unclaimed.
The index is the answer to "who owns this category in AI search, and where is nobody?"
Why open with it
Three reasons this is a stronger first deliverable than a brand audit.
But before the reasons, the tension. AI search is not a gradual shift; it is a category-locking event. The brands cited in the first eighteen months of an LLM's answers are the brands the model returns to forever, because every fine-tune builds on the last set of citations. If your client is invisible while their two biggest competitors get listed in 40% of category prompts, the gap compounds week over week. It does not wait for them to publish a content calendar.
The business impact is concrete. Gartner has projected a 25% drop in traditional search traffic by 2026, and the prompts that drive purchase decisions are migrating off Google fastest. "Best [category] tool for [use case]" is now routinely asked of an LLM, especially by technical evaluators and younger buyers who default to ChatGPT before they default to Google. If your client is not in that answer, the lead never reaches their site. There is no ranking page to fix, no SERP feature to optimise. They are simply not in the conversation.
The Visibility Index makes that visible in one document. Without it, the client's complaint is "we should rank higher." With it, the conversation reframes to "the category has three winners and we are not one of them. What does it cost us to change that, and how fast." That is the conversation that closes engagements.
It is non-threatening. A brand audit walks in with a list of the client's gaps. Useful, but it puts them on the back foot before you have built any trust. The index walks in with a map of their entire category. The client looks at it and locates themselves; you do not have to point at the holes.
It exposes the real conversation. Most prospects think their problem is "we are not cited enough." The index usually shows the actual problem is structural: the category is concentrated around two or three players, the LLMs lean on three or four domains, and the prospect appears in none of them. That is a different conversation, and a more honest one.
It scales to prospects who have not bought yet. A brand audit needs the client's brand. The index needs only an industry and a topic. You can run one before the discovery call, walk in with the map, and use it as the meeting itself, not as homework after the meeting.
How to run one
The fastest path is the MCP server in Claude Code. Install it once (see the previous post), then this prompt creates a working index in a few minutes:
> Use the AEO Copilot MCP to build an industry index for "Project management for engineering teams." Create three topic clusters: tool comparisons, JTBD evaluation, and pricing. Add 5 prompts per topic that an engineering manager would actually ask. Run all prompts across the four LLMs. When the run is done, pull share of voice, top sources, and whitespace, and summarise what the category looks like.
What it does, in order:
create_index (name + industry + description)- Three
add_index_topic calls
- Three
add_index_prompts calls (5 prompts each = 15 prompts × 4 LLMs = 60 runs)
run_index_prompts to fire them allget_index_share_of_voice, get_index_sources, get_index_whitespace for the analysis
The whole loop runs unattended.
Reading the output
A real index for a real category looks something like this. Numbers are illustrative, the shape is not.
Share of voice
Concentration score: top-1 share 41%, HHI 2,431.
The HHI tells you how monopolised the category is. Anything over 2,500 is a market the LLMs treat as dominated by one or two players. Anything under 1,500 is fragmented; lots of room. In between is contested.
For the client, the takeaway is not "you are not on this list." It is "this category is concentrated; getting onto it requires beating Linear at the source level, not just publishing more content."
Sources
This is where the LLMs are getting their answers. If your client is not represented in these domains, no amount of on-site optimisation will move them. The action item writes itself: get covered on Reddit, get reviewed on G2, get a Hacker News thread. The agency conversation that follows is about distribution, not just content.
Whitespace
Whitespace is filtered to prompts where no entity appears in at least 50% of runs. These are the gaps. For each one, an entity that publishes consistently and gets cited two or three times in a row will start to lock in.
This is the most actionable view in the index. It is also the one that closes the meeting: "Here are five prompts in your category that nobody owns. Pick the three closest to your positioning and we will go after them."
Use it as your top-of-funnel
The index is also why this stops being a sales asset and starts being a lead-generation engine.
Pick an industry you want clients in. Run an index for it. The output is a short brief: who is winning, who is cited, where the gaps are. Send that brief to ten companies in the category, cold, not from your existing pipeline.
What you are sending is not a pitch. It is a public-good document about their category that they could not have made themselves. The reply rate on that is materially different from generic outreach. The agencies I have shared this workflow with run one index per target industry per month and use the resulting briefs as their entire top-of-funnel.
Three things compound:
- Authority position. You are the person who maps categories. That is a different kind of vendor than the one selling content packages.
- A useful artefact, not a pitch. The brief names specific prompts, specific competitors, specific gaps. It is useful to the recipient before they buy anything.
- An open door. "I ran an index for [industry], here is what I found" is a meeting opener that does not need permission first.
How to use it in a discovery call
The structure I use, every time:
- Open with the share-of-voice table. Let the client read it. Ask them where they think they should be. They will tell you what they actually believe about their own positioning, which is more useful than asking directly.
- Show the sources. This is usually where the room shifts. Most clients have not thought about LLM citations as a distribution problem; the sources view makes it concrete.
- End on whitespace. Give them three prompts, no entity owns them, and explain what it would take to claim one. This is the deliverable they remember.
- Close with a brand audit as the next step. Now you have permission to look at their specific gaps, because you have already shown them the category.
The index is not the work. It is the door-opener that makes the work obvious.
What it costs to run
The index uses the same prompt budget as your normal monthly limits. Fifteen prompts across four LLMs is sixty runs. If you are on the Agency plan that is a small fraction of the monthly allowance. The marginal cost of running one for a prospect, in other words, is essentially zero, while the conversion lift on first calls has been the highest single change I have made to my own pitch.
One catch worth noting up front: indexes are currently API-only. You need a plan with API access, which today means the Freelancer or Agency tier. The prompt budget shares your normal monthly allowance, so running 60 prompts a month for an index does not eat into your brand-tracking budget noticeably.
Try it
Indexes are API-only for now. There is no dashboard UI for them yet. If you have a Freelancer or Agency plan, you have access; the MCP server in Claude Code is the simplest way to drive it.
If you do not have an account: start a trial, pick one industry you want to win, build the index, and walk into your next discovery call with the map.
Get your API key →