Working notes on international SEO and AI search visibility — checklists I actually use and opinions I am willing to defend. One substantial piece a month, each carrying something you could not get from a generic SEO blog.
Who holds the top ten for treatment searches in the US, Germany and Spain: institutions take 72% of US results. Where a health brand can compete, and a six-step plan.
Manufacturers hold 13% of top-10 results for treatment searches in Germany and 2% in the US. The law behind it, and what 43 pharma and medtech sites allow AI crawlers.
Five decisions, and the one nobody reviews: how the site routes visitors between languages. What 110 international brands do, and the two that redirect their own language pages.
The same query needs a median of 111 linking sites to rank in the US and 8 in Germany. Where local links come from, and how to split a budget by market.
Germans search Siebträgermaschine, not Espressomaschine. One product in four markets, what buyers expect in their language, and when translating is enough.
Each engine runs two or three crawlers, and blocking the wrong one takes you out of the answer entirely. What the vendors document, and what the robots.txt of 230 brand sites actually allows.
How hreflang works, where it goes, and how it interacts with canonicals — Google’s rules, two large error studies, and what 50 international brand home page sets got wrong in September 2026.
Search Console now reports AI Overview and AI Mode impressions, but no clicks and no queries. What each data source can tell you, where rank trackers mislead, and the monthly routine I run.
Spain searches “alquiler de coches” 26,000 times a month, Mexico 200. Search volumes for one service in five markets, what Keyword Planner and Google Trends cannot compare across countries, and the process I use instead.
What x-default tells Google, the case it was designed for, the mistakes I find in audits, and what 19 international brand home pages actually do with it.
Every AI-visibility checklist says to add one. The server logs of 137,210 domains say almost nothing reads it. What the evidence shows, why I publish one anyway, and what to spend the ten minutes on instead.
What Google’s documentation actually says about international URL structures, the costs its comparison table leaves out, and the decision rule I use for multi-market sites.
Your second market underperforms and the technical audit came back clean. The five-dimension matrix that explains it in an afternoon — and why the fix is rarely more translation, and sometimes fewer pages.
Translation optimises for meaning; SEO localisation optimises for how people search. The eight things a translator cannot know from your source page, and the brief that closes the gap.
The usual first EU market and the most often underestimated. The competitive set you did not analyse, compound nouns that break translated keyword research, the trust signals German buyers look for, and the imprint German law actually requires.
Ranking first does not mean being cited. How engines actually choose sources, why they answer the same question differently in every language, and why your smallest market is the one to win first.
The cost comparison against an agency and an in-house hire, stated plainly — and the three situations where a fractional lead is the wrong answer, which most articles on the subject leave out.
Architecture, hreflang, content parity, per-market intent, geotargeting and measurement — the checks I run on a multi-market site, in the order I run them, with the three findings that account for most of the damage.
What changes when health content crosses a border: YMYL scrutiny, medical review as a scheduling constraint rather than a checkbox, claims that do not survive a border, and E-E-A-T a machine can actually read.
One substantial piece a month rather than a stream of short posts. Each one carries either an original checklist, a dataset, or numbers from my own site — if a note has nothing in it you could not get from any other SEO blog, it does not get published.
No client case studies yet. I cannot publish my employer's data, and I will not write a case study about a client without their written permission. When there is one, it will appear here with real figures and a named client, or not at all.