Case Study: Getting a Bilingual Cyprus and Greece B2B Site Cited by AI Engines
A B2B business serving both Cyprus and Greece arrived at a familiar crossroads. Google rankings still mattered, in two languages, but a bigger shift was underway: buyers were increasingly asking AI assistants like ChatGPT, Perplexity and Google’s AI Overviews for a recommendation before they ever opened a results page. The goal was no longer just to rank. It was to be the answer. Here is how we approached it.
What we found
The English foundations were healthy: unique titles and descriptions, a clean sitemap, FAQ schema, a proper canonical setup. The problems sat in the places that decide whether a bilingual site gets found by machines.
- The Greek pages declared themselves as English in their language attribute, and there were no hreflang tags, so search engines risked treating the two languages as duplicates and serving the wrong one.
- The pricing an AI agent could read did not match the pricing on the page. The structured data still described an older set of packages.
- There was plenty of content, but it was written to be read top to bottom, not in the answer-first way AI engines extract and cite.
- The writing carried the small tells that make text read as machine-generated, which quietly erodes the trust that earns citations.
What we changed
Fixed the bilingual foundation
Every page now declares its real language, and reciprocal hreflang links connect each English page to its Greek counterpart. Visitors and search engines get the right language, and the two versions stop competing with each other.
Made the business readable by AI agents
We published a machine-readable pricing file so an AI assistant comparing providers can parse the plans without rendering a single line of JavaScript, and we made the pricing identical across the page, the structured data, and the contact forms. When an agent can read you, it can recommend you.
Restructured content for extraction
AI engines cite passages, not whole pages. So the key pages and articles now open with a self-contained answer of forty to sixty words, use comparison tables where buyers weigh their options, and carry FAQ blocks. These are the formats answer engines lift most often.
Built the content the buyers were about to search
We produced a clear, sourced guide to the EU AI Act’s transparency rules for chatbots, tied to a hard deadline, in both English and Greek. It is exactly the question local businesses were starting to type into a search box, and almost nobody was answering it in Greek.
Made the writing sound human
We removed the punctuation habits that read as AI-written across the whole site, so the content sounds like a person wrote it. That is not cosmetic. Trustworthy, human writing is what both readers and answer engines reward.
Set up the local signals
We added proper local-business structured data with real coordinates and service areas, made the name, address and phone consistent everywhere, and mapped the Cyprus and Greek directories worth being listed in.
The principle behind it
The instinct when you want more visibility is to publish more. That was not the problem here. The content already existed. The work was making it extractable, consistent, and trustworthy, in both languages, so that a machine reading the site could understand who this business serves, what it offers, and why it should be the recommendation. Structure beats volume.
Where it stands now
The site serves the correct language to every visitor, exposes machine-readable pricing to the AI agents that increasingly shortlist providers, leads its important pages with a citable answer, and owns the one compliance topic its buyers are actively searching, in English and Greek. The foundation to be cited, not just ranked, is in place.
Want your site found and recommended by AI engines, in one language or two? Book a free audit and we will show you exactly where the gaps are.
