Most multilingual SEO programmes are bolted-on translations of a single-language site. The result: confused search engines, fractured entity signals and AI assistants that recognise your brand in English but not in Serbian, German or anything else. I work the other way around — one entity, many languages, engineered for both classical and AI-driven search.
— (01) The discipline
Multilingual SEO is the discipline of making a single brand discoverable, trustworthy and citable in every language and market it operates in — without diluting itself in any of them.
The work covers international SEO (geography), language-level SEO (translation, localization, market intent), and the increasingly important multilingual entity layer — making sure ChatGPT, Perplexity, Gemini and Google AI Overviews recognise your brand as one entity, regardless of the language of the question.
Done well, multilingual SEO is invisible to users and unmissable to search engines. Done poorly, it duplicates content, breaks hreflang, fragments authority and leaves your strongest market subsidising your weakest.
— (02) Markets & languages
A focused regional practice — strategy and execution in three languages, review fluency in two more. For other markets I lead the strategy and partner with vetted native specialists.
Native Serbian. Cyrillic and Latin scripts. Local SERP, ekapija, Google.rs and AI assistants in regional context.
Default working language for strategy, deliverables and global ProfessionalService positioning.
Working German. Market-specific keyword research and entity coherence in Google.de / .at / .ch.
Read-and-review fluency. RTL technical considerations, Hebrew entity signals, kosher hreflang.
Read-and-review fluency. Yandex-aware strategy where it still matters; Russian entity coherence in LLMs.
Strategy lead with vetted native specialists. The frameworks travel; the language layer doesn't need to be mine.
— (03) Services
Engagements compose. Most clients begin with the international SEO audit and entity work, then layer the rest as the programme matures.
The map before the build.
A full audit of how your domain, sub-folders, sub-domains, hreflang, canonicals and structured data behave across markets. Output: a prioritised plan that names the 20% of issues responsible for 80% of leakage.
Get the plumbing right.
hreflang implementation (HTML, sitemaps, headers), URL strategy (ccTLD vs sub-folder vs sub-domain), canonical hygiene, geo-targeting in Search Console, and redirect logic that actually preserves equity.
One brand, every language.
Wikipedia / Wikidata signal hygiene per language, sameAs networks, market-specific schema (Person, Organization, ProfessionalService), and the entity coherence work that lets a single LLM concept resolve correctly across every locale.
Translation isn't localisation.
Information architecture, page templates and editorial guidance built for each market — including locale-specific keyword research, intent mapping and answer-block patterns. Translation memory ≠ market intent.
Cited in every language.
GEO and AEO work scoped per market. Citation engineering, source-worthy content frameworks and AI Overview tracking — measured in each language so a strong showing in English doesn't mask weakness elsewhere.
Skill the people who'll keep doing it.
Workshops and embedded coaching for in-house SEO, content and product teams operating across multiple languages — delivered in English, Serbian or German depending on the team.
— (04) Symptoms
The recurring patterns from cross-border audits over the last decade.
One language carries the brand.You rank in English. Everywhere else you're a translated PDF.
Broken hreflang.Versions self-canonicalise, point at deleted pages, or contradict the sitemap.
Translated, not localised.The German pages read as English in disguise. Google notices. Buyers do too.
Fragmented entity.LLMs treat your English brand and your Serbian brand as two different (and weaker) entities.
Schema in the wrong language.Structured data hard-coded in English on non-English pages. Confusing every retriever.
No multilingual AI visibility plan.You have an SEO programme. It stops the moment AI Overviews and ChatGPT are involved.
— (05) Why work with me
I'm an independent AI & multilingual SEO consultant based in Belgrade. The practice has been running since 2013, intentionally small — fewer than eight clients per year.
What clients hire me for: strategic clarity in three working languages, reading fluency in two more, and a habit of treating multilingual SEO as one continuous entity problem rather than a translation backlog. I write the briefs, design the schemas, supervise the implementation and personally review hreflang, canonical and entity decisions across markets.
What I refuse to deliver: bulk machine translations dressed as SEO, locale-blind link buying, or strategy decks that look the same in every country except for the flag.
— (06) Framework
Every multilingual engagement runs through the same disciplined sequence — adapted per language, but never skipped.
Crawl per locale. Map current visibility, hreflang health, canonical conflicts, schema language, market-level performance and AI search behaviour.
Define the one entity behind every locale. Wikidata, sameAs, Person and Organization graphs aligned across languages. Disambiguate where the brand collides with another.
URL strategy, sub-folder vs ccTLD vs sub-domain, hreflang implementation (HTML, sitemap, headers) and the canonical hygiene that keeps each version visible.
Per-market keyword research, intent mapping, answer-block patterns, editorial briefs. Translation memory becomes a starting point, not a finish line.
JSON-LD in the right language per page. FAQ, HowTo, Article and Service schemas adapted to each market. Citation outreach where it earns measurable lift.
Dashboards split by language: organic, AI Overview share, LLM citations, branded prompt accuracy. We iterate weekly per market, not in one undifferentiated bucket.
— (07) Outcomes
Outcomes are measured per language, against pre-engagement baselines. No single-language vanity wins.
Not just the strongest market. Each locale moves on the metrics that matter to it.
Wikidata, Knowledge Graph and downstream LLMs resolve to a single concept, regardless of the language of the query.
No more self-cannibalising language versions. Search Console international report stays green.
Cited and recommended in ChatGPT, Perplexity, Gemini and Google AI Overviews — in each target language.
A repeatable playbook for opening the next language without rebuilding the strategy each time.
The people responsible day-to-day know how to run it after the engagement ends.
— (08) FAQ
Eight short answers. Anything else, ask in the call.
— Ready to expand cleanly
If you're running multilingual SEO that feels held together with tape — or you're about to open a new market and want to do it right the first time — let's talk. Booking Q3 / Q4 2026.