AI search methodology
Our AI Visibility Methodology
Five phases, one locked prompt universe, and two scores per prompt — citation and recommendation. Everything we publish about your program traces back to this framework.
The five-phase framework
Phase 1
Discover
We build the prompt universe: 100+ real buyer questions derived from your buyer research, sales-call language, and named competitor set. Nothing downstream is trustworthy if the prompts aren't the ones your buyers actually type.
Phase 2
Map
We baseline every prompt across ChatGPT, Google AI Overviews, Gemini, Perplexity and Copilot, scoring citation and recommendation separately, then map the source ecosystem — the third-party sites the engines actually pull from in your category.
Phase 3
Optimize
Answer-first page structure, entity clarity, schema and structured data, and retrieval-friendly technical foundations. The goal is to make your content the easiest correct source for a model to lift.
Phase 4
Build Authority
Earned coverage, expert commentary, and Gap-Fill Placement on the specific third-party sources the mapping phase proved the engines rely on — never volume placement for its own sake.
Phase 5
Measure
The prompt universe is locked and re-run on a fixed cadence. Monthly reporting shows citation movement and recommendation movement as separate lines, with a quarterly competitor benchmark.
The seven prompt-intent types
Every prompt universe is built from these seven intents so coverage spans the full buying journey, from unaware category discovery to late-stage pricing checks.
- 01
Category discovery
"What tools exist for X?" — the buyer doesn't yet know the vendor landscape.
- 02
Comparison
"A vs B" — two named options are already on the shortlist.
- 03
Alternatives
"Alternatives to A" — an incumbent exists and is being displaced.
- 04
Problem-led
The buyer describes a symptom, not a product category, and asks how to solve it.
- 05
Constraint-led
Requirements-first prompts: compliance, region, integrations, team size, budget ceiling.
- 06
Validation
"Is A any good?" — a decision is nearly made and the engine is being used as a reference check.
- 07
Pricing
Cost, licensing, and total-spend prompts, where published numbers strongly influence the answer.
Methodology FAQ
What is the difference between an AI citation and an AI recommendation?
A citation means the engine used your page as a source. A recommendation means the answer names your brand as the option to consider. Research shows a cited source is not the recommended brand roughly 43% of the time, which is why we score and report both separately.
How large is the prompt universe?
Every engagement starts with 100+ prompts built across seven intent types: category discovery, comparison, alternatives, problem-led, constraint-led, validation, and pricing. The universe is locked and versioned so month-over-month movement measures program performance, not prompt drift.
Which AI engines do you measure?
ChatGPT, Google AI Overviews, Gemini, Perplexity and Copilot — the five engines where B2B buyers most commonly run vendor research.
Ranking is no longer enough
You need to be cited, mentioned, and recommended.
Get a free AI Visibility Report — see exactly where your brand appears across ChatGPT, Google AI, Gemini, Perplexity, and Copilot, and where competitors are winning instead.