International SEO and GEO · the Canopée plan
Every market has its own engines. And its own answers.
A multilingual site is not translated, it is structured: hreflang, market-level architecture, local intent. And since 2026, AI engines that do not cite the same sources in Paris, Berlin and Madrid. We measure both and work both, market by market.
The same question, three different answers
The content exists, nobody picks it up. Your site is translated, but no local source cites it. The engines have nothing to quote in the market's language.
Cited, but through a single door. Your brand appears via one local comparison site or directory. The citation exists. The preference does not.
The position you defend. Several local sources converge and the engine recommends you. It is the only state you defend instead of conquering.
"Best accounting software" does not get the same answer in French, German and Spanish: different sources, different brands, different competition. Inside one country, engines split: in Botanik's Baromètre GEO France, the 3 engines disagreed on the number 1 in 2 of 4 markets. Our audit returns one verdict per market: Absent, Mentioned or Recommended, engine by engine, language by language.
Six work streams that only exist across borders
ccTLD, subfolders or subdomains. The right structure depends on your actual resources and maintenance capacity, not on doctrine. We decide per market and document why.
Return links, canonicals, consistency between declared and served languages. Hreflang is the first source of errors on multilingual sites. We audit it pair by pair.
You do not translate keywords. Each market has its own queries, its own seasonality, its own comparison sites. We rebuild the demand map per country.
AI engines cite local sources: press, directories, reference sites. Being cited in France says nothing about Germany. The source work restarts country by country.
The Absent / Mentioned / Recommended verdict, measured per engine and per language, with a baseline per market. The full spec is published on the Oracle method page.
One consolidated view for headquarters, autonomous local teams for execution. The model built for store networks, applied to countries.
International SEO adds six work streams that a single-market site never faces: market architecture, hreflang, local intent, sources per language, per-market measurement and central governance. Each one is decided on your real resources and your real demand data, not on dogma, and each decision is written down.
Why AI engines disagree from one market to the next
AI engines do not share an index. ChatGPT leans on its own corpus and Bing search; Gemini leans on Google's index and its SEO signals. In 2026, 38% of Google top-10 sources are cited in AI Overviews, against 76% in 2025. Different substrate, different answer, in every language.
International GEO is the work of measuring and improving how AI engines cite your brand across several markets and languages, one verdict per market: Absent, Mentioned or Recommended. This page covers that cross-border layer. What a GEO agency does end to end is covered on our GEO agency page (/en/generative-engine-optimization-agency).
The substrate engines quote from is local. The comparison sites, trade press and directories an engine cites in Spanish are not the ones it cites in German. A brand that ranks on Google in several countries can still be Absent from AI answers in most of them, because no local source gives the engine anything to quote in that language.
The numbers argue for structuring now, not later. AI-driven queries grew 494% between 2023 and 2026. Google still records 84 billion visits per month against 5.5 billion for ChatGPT, so classic search carries the traffic while AI answers shape the shortlist. International work has to serve both: rankings per country, citations per language.
How each engine is interrogated, scored and thresholded is not this page's job. The full specification is published, version number and limits included, on the Oracle method page (/en/oracle-method).
How a multi-market engagement runs
A multi-market engagement follows four stages, and each market gets its own baseline before anything is optimized. Measurement follows the published Oracle method: up to 5,000 demand-weighted prompts per market, replayed on each engine. The stages below are what changes when several markets are in play.
- 01Sequence the marketsWe rank your markets on real demand data: query volumes, competitive density, existing assets. You do not launch six countries at once. You launch where the demand map says the return is.Deliverable : A market sequence with the data behind it
- 02Set the baseline per marketThe Oracle audit replays the market's prompt corpus on each engine, in the market's language, and returns your verdict: Absent, Mentioned or Recommended, engine by engine. The baseline is signed and measured by a third-party tool before any action starts.Deliverable : A signed baseline per market, per engine
- 03Fix the structure, then the sourcesArchitecture and hreflang first, so every version is crawlable and attributed to the right market. Then the citation work, language by language: local press, comparison sites, directories, the sources engines actually quote in that country.Deliverable : A shipped technical base and a per-country source plan
- 04Report per market, decide per marketMonthly reporting shows each market against its own baseline, engine by engine. A market that moves gets more investment. A market that stalls gets a documented decision: adjust or stop.Deliverable : A monthly per-market report against baseline
What does international GEO cost?
Canopée, Botanik's multi-market plan, starts at €2,670 per month excluding VAT: at least 3 days of senior time each month, at the single €890 day rate that applies to every client and every volume. The name is French for canopy, the top layer of the Botanik range.
Canopée includes everything in Pousse, our growth plan (Pousse is French for shoot), deployed across several markets, plus zone-by-zone geogrid measurement for local visibility, PR and citations, and a dedicated mission lead.
If you want the diagnostic first, Graine (French for seed) is the audit tier: from €2,670 one-shot, delivered in 3 weeks. Continue into Pousse or Canopée within 60 days and the audit fee is deducted from your first month of support.
Plans run on 6-month engagements for a reason we publish: engines, classic and AI alike, take 3 to 4 months to re-crawl, re-cite and re-rank. Below 6 months, you pay to sow without ever harvesting.
The full grid, the plan floors, the 15 to 25% performance-indexed share of our fees and the 2026 market price ranges are published on our GEO cost guide (/en/what-a-geo-agency-costs). We do not do quote-only pricing.
The limits we publish before you sign
The Oracle spec ships four limits with every report, and one matters most across borders: French prompts dominate the corpus today, and multilingual coverage is partial. We write that in the report rather than sell around it, and each market's corpus is built from that market's own demand data.
The other limits hold internationally too. Five runs per prompt reduce noise, they do not cancel it. An engine can cite an old state of your market, so every collection is dated. And engines change without notice, so comparisons only hold at constant methodology, which is why the method carries a version number.
One consequence you will see in your reporting: a movement between two measurements only counts as significant when it clears the sum of the two stability bands. Below that, it is noise, and the report says so.
A French agency that operates in English
Botanik is a French agency that runs its own site in two languages, French and English. The page you are reading is the English twin of a French original, maintained in production alongside it. We sell multilingual structure because we operate one ourselves.
The same goes for people. Your markets are handled by seniors only: no juniors, no subcontracting, and the person in front of you is the person doing the work.
And for reversibility: your content, your access and your data belong to you on every market. If you leave, everything follows you.
Why is translation not enough for international SEO?
Because each market has its own ranking criteria on Google and on other engines such as Baidu or Yandex, its own queries and its own seasonality. A translated page with no local structure and no local citations can stay invisible in classic results and in AI answers alike.
How do you structure a site for several languages?
With hreflang tags and the right redirects, to signal to Google which version to serve in each country, and with an architecture decided per market: ccTLD, subfolders or subdomains depending on your real resources. Return links and canonicals are checked pair by pair; they are the first source of multilingual errors.
Do AI engines really answer differently per country?
Yes. In our Baromètre GEO France, the 3 engines disagreed on the number 1 in 2 of 4 published markets, inside a single country. Across languages the gap widens, because each engine quotes local sources: press, directories and comparison sites specific to the country.
How long before AI engines cite us in a new market?
First citations move in 6 to 10 weeks. A durable shift from Mentioned to Recommended plays out over 4 to 6 months, which is why support runs on 6-month engagements.
Which engines do you measure?
ChatGPT, Perplexity and Google AI, interrogated under equal conditions: same prompts, same conditions, via API, each engine measured separately. The engine list, collection protocol and scoring live in the published Oracle spec, version included, on the Oracle method page (/en/oracle-method).
Can we start with a single market?
Yes. Graine, the audit tier, starts at €2,670 one-shot and is delivered in 3 weeks: your verdict per engine, a full SEO audit and a prioritized roadmap. If you continue into Pousse or Canopée within 60 days, the audit fee is deducted from your first month.
Start by knowing where you stand, market by market.
The international audit sets a baseline per country: positions, traffic and your AI verdict per engine and per language. That is the scope of the Canopée plan, priced in the open like everything else, from €2,670 per month. Want a first signal today? The free scan returns a verdict per engine in 60 seconds, no account, no card.
Book the audit, or run the scan first. Either way, you will know which of your markets defend a position and which still have to win one.
Case studies, measured results
How Pennylane strengthened its SEO presence on key accounting and financial queries
With Botanik, Pennylane structured a comprehensive SEO strategy combining technical optimization, educational content, and semantic targeting to strengthen its position in the financial management tools ecosystem.
Franck Neuenschwander · founder, author of the Oracle method
The method you just read, Organic Inference™, is mine: published, versioned, with its limits in writing. Fourteen years on the advertiser side, and the only agency I have ever worked at is the one I run: if you want to discuss it, you discuss it with me, not with a salesperson.
Thirty minutes with a consultant
Not a salesperson. We tell you which lever fits your case, and if none does, we say that too.
Are you absent, mentioned or recommended?