Services
Paid analysis and implementation for a specific brand, category and decision.
Research
Open research on AI sources and recommendations.
This is where methods, research notes and, later, versioned data are published. Every release names its measurement surfaces, prompt population, denominator and limits. Client samples and open panels remain separate.
Four areas, four roles.
Research
Open methods, research notes, benchmarks and publishable data.
Case Studies
Anonymised results from client work, with project-specific measurement boundaries.
Lab
Runnable tools, workflows and skills, including the DIY Share of Voice monitor.
In development
DACH AI Source Benchmark 2026
The planned project is an open fixed-panel study for Germany, Austria and German-speaking Switzerland. Before the first public baseline, the prompt panel, measurement configurations, method and publication rules will be frozen.
Until at least three comparable waves exist, the project is called a benchmark, not a monitor. No trend or population-representative claims are being published yet.
Research Note 0 · Client-derived sample
Why visibility in one AI engine says little about the next
One month of retrieval-to-citation data from eight anonymised B2B projects shows how strongly observed source pools and citation rates differ by engine.
Data boundary: this is an observational analysis from client work, not the controlled public DACH panel. Aggregate tables are open; project-identifying raw data remains protected.
What counts as evidence here.
Mention, consideration, recommendation and citation are separate measurement signals. “Source” means an attribution exposed in the measured response, not automatically a hidden retrieval or training source. Every percentage states its denominator.