Approach
How I plan, measure and run SEO work, and the tools I use for it.
Semantic SEO and entity architecture
I plan a site around the entities it needs to be understood for: the organisation, its services, places and people, and how they relate. That model drives the topical map, the internal linking and the structured data, so all three describe the same thing.
On multi-market work the entity layer is shared and the content is localised per market. The organisation and service definitions stay the same; regulation, terminology and proof points change by country.
The approach has limits. Topical coverage built mostly from informational articles is the layer that core updates and AI Overviews have hit hardest. The multi-market case study is partly about that.
Measuring AI visibility
Search Console doesn’t show whether a brand is mentioned or cited in AI answers. I track that separately in Peec and Profound: a fixed set of prompts per market, run on a schedule, recording whether the brand is mentioned, whether it’s cited, and which sources are cited instead.
Read next to Search Console, it separates two things that look the same on a traffic chart: losing rankings, and keeping rankings while the answer moves onto the results page.
Audits and research in Claude Code
I run audits, crawls and research through Claude Code connected to Ahrefs and DataForSEO over MCP, rather than through the tools’ interfaces. A question like “which pages earn organic traffic but are missing from the sitemap or return a redirect?” becomes one working session instead of several exports and a spreadsheet.
The steps are written down as they happen, so the same analysis can be rerun next month and compared.
Building tools when work repeats
When a manual process comes up again, I script it instead of doing it a third time. This site follows the same standard: static HTML, structured data on every page, and nothing that needs JavaScript to read. View the source or run Lighthouse on it.