Before you write anything, find out what the AI already believes about you.
Mapping is the first phase. It answers four questions with evidence, not assumption: what the AI already believes about your brand with no search involved, which competitors it keeps naming no matter who you think you compete with, which sources it cites in your category, and which questions your buyers actually ask. Every decision after this is priced off these answers.
Map is the first phase of the Citabld method. Citabld records what AI tools already say about a brand, how strongly they recommend it, which competitors they name alongside it, which sources they cite in the category, and which questions buyers actually ask. These findings become the inputs for the question library.
What is a parametric baseline?
A parametric baseline is a record of what the AI says about your brand and category when no search happens at all, drawn from what it was trained on. We probe it on purpose, including on tools that don't search by default, because that trained picture shapes every awareness and consideration answer, and you can't influence a search that isn't happening.
What we capture
How each tool describes you unprompted, which details it attaches, which competitors it names with you, and where it's simply wrong about facts we can fix at the source.
Why it matters
Early-stage topics are measured on whether you're included and how you're described, never on citations. The baseline is the only starting line those numbers can be read against.
What is a prominence tier?
A prominence tier is our read of how strongly the AI already recommends you, set through a live probe. Tier sets the depth of everything after it, because buyer-type depth your market position can't use is money spent on nothing.
| tier | how the AI behaves | what we build |
|---|---|---|
| leader | recommended across nearly every buyer type | light buyer-type work, defend the set it already names |
| mid market | the names swap as the buyer type changes | the full buyer-type matrix, maximum value |
| long tail | rarely named without a search | visibility work first, buyer-type depth later |
If the AI doesn't know what you are, nothing downstream can work.
Mapping ends with a written verdict on who you are to the AI, your visibility tier, and every buyer question you gave us. Some of those verdicts say do not build here.
How do you source head queries?
Buyer questions come from real buyer language: sales calls, support tickets, win and loss notes, and the communities your buyers post in. Keyword tools produce keyword-shaped phrasing, which produces keyword-shaped fan-outs, which miss how people actually talk to an AI. Every question you give us gets a written verdict, so nothing is silently dropped.
| verdict | meaning |
|---|---|
| accept | the query is how buyers really ask it; it goes into the library as supplied |
| reshape | the intent is right, the phrasing is not the AI's vocabulary; we rewrite and show both |
| stage correct | the query sits at a stage where the metric you want does not exist; we retag it |
| split | the query bundles two intents that fan out differently; we separate them |
| park | no demand evidence found; we park it and show you the evidence we looked for |
What does the citation landscape show?
The citation landscape is a live capture of which sources the AI actually cites for your category's questions, by question type. Comparison questions pull heavily from third-party sites, and the roundups that dominate them are often competitors' own list posts. Local and branded questions pull heavily from your own pages and listings. This capture is what turns where you publish into a scored decision instead of a default to your blog.
- /Which publishers and review platforms appear repeatedly in your category.
- /Which competitor-owned pages are being cited in answers about you.
- /Which surfaces are institution-owned and therefore not worth fighting.
- /Where your own domain already earns citations and should be reinforced.
See this phase run on one of your own head queries, free.
Six questions, one modeled fan-out across seven engines, and one finished piece of content for the subquery you pick.