AI Search
You cannot fix a channel you have never measured.
Find out when AI recommends you, when it recommends your competitors - and why.
The audit is the entry point to everything else here. We build a prompt universe for your category, run it against the major models, record the full answers together with the sources they used, and report where you stand.
The output is not a score. It is a list of specific answers where a decision was made about your category, showing who was named, on what basis, and what would have to be true for you to be named instead.
What it covers
- Prompt universe built from real customer questions
- Measurement across five engines, split by market
- AI Share of Voice and competitor benchmarking
- Citation and source map for your category
- Site readiness: what stops models from quoting you
- Roadmap with the work ordered by expected effect
Also in this capability
- GEOGenerative Engine Optimization: the work that makes your content usable as an answer.
- LLM OptimizationMake your site readable, retrievable and safe to quote for language models.
- AI Brand MonitoringContinuous measurement of what models say about you, in which context and on what basis.
- Entity OptimizationRemove every doubt about what your brand is, what it does and what it is connected to.
How it works
01
Build the prompt set from how people actually ask
Not keywords. People write whole situations to a model - budget, constraints, what they tried already. The prompt set comes from sales calls and support tickets, because that is where those sentences exist in their original form.
02
Separate general from brand visibility before reporting anything
A prompt containing your name will produce a mention almost every time, including when the model is saying it knows nothing about you. Mixed together, those two numbers produce a score that rises whenever you add prompts with your own name in them. We count them apart from the first run.
03
Record the sources, not just the answer
Every answer is stored with the pages the model used. That turns 'we are not visible' into a list of specific publications that decide your category - which is the difference between a finding and a plan.
04
Say what would have to change
For each prompt where a competitor is named, we state what is missing: a fact that is nowhere on your site, a contradiction the model would not risk, or a source you are simply absent from.
What you get
- Prompt set built from real customer questions, reusable for later runs
- Visibility split into general and brand, reported separately
- Competitor list as the models see it - usually not the one you expect
- Source map: which publications decide answers in your category
- Site readiness findings: what stops a model from quoting you
- Roadmap ordered by expected effect on the next measurement