Case studies
Client work and own research, each written as problem, diagnosis, implementation, measurement and result. Clients are not named. Every figure carries the source it came from.
Lead case
A UK legal practice
A brochure site on a website builder, rebuilt into fifteen structured pages across three languages, then measured for eight months with search data and server logs kept separate.
Client projects
A multilingual medical website
Technical and content optimisation in a category where trust signals decide whether a page is used as a source.
A personal care e-commerce site
Category, product and FAQ architecture, with product facts written as text rather than left in the design.
A second legal property
The same firm's international site, treated as a separate engagement with its own measurement period. Published when the reporting window closes.
Own projects and research
Not client work. This is where the methods are built and tested before they reach a client site.
Pensne.lt
A multilingual asset used to test semantic structure, entity signals and AI visibility under controlled conditions.
Pensne Digital
Bot access, structured data, technical health, on-page foundations, and how citation checks are run.
Your site, same treatment
Send a domain and I will check what AI systems can reach, read and retrieve.
The Articles
Why AI Answers Skip Real Experts
Generative engines do not rank expertise. They retrieve passages that are clearly attributed, structurally clean, and confirmed on more than one domain. A specialist with fifteen years of real practice can be invisible to those systems, while a competitor with half the experience and twice the structure gets cited by name.