This blog publishes every post in Spanish, English, and Japanese. They are not translations. They are three versions of the same topic written for different readers.
The difference matters. A translated post says the same words in another language. An adapted post says the same ideas with the words, structure, and emphasis each audience needs. Translating is changing the language. Adapting is changing the context.
Why I do not translate
I started by translating. The first posts I wrote in Spanish and then converted to English and Japanese with Claude's help. The result was grammatically correct and culturally incorrect.
A paragraph that in Spanish opens with personal context ("I have been designing for banks for 10 years") works because, in my experience, the LATAM audience I write for responds better to personal history. The same paragraph in English sounds like self-promotion. In English I have found that data and results work better as an entry point. In Japanese I prioritize technical precision and state-of-the-art context.
When I stopped translating and started treating each language as its own text, quality went up in all three. Not because the words were better. Because the structure responded to what each reader expects.
The process
I write the concept once. The topic, the data, the findings, the section structure. That is universal. An experiment where 4 AIs predicted my answers and all 4 built an idealized version of me is the same finding in any language.
Then I generate the three versions with Claude in the same session. Not one after another. All three together. Claude receives the concept, the data, and the style guide with the rules per language. It produces all three versions knowing each one is independent.
Then I review. The Spanish version I review directly. The English version I review for tone and structure. The Japanese version I verify against the technical concepts that must be precise, but I trust Claude more for the naturalness of the language.
The result is three texts that share the same thesis but do not share the same sentences.
What changes between languages
It is not just the words. Four things change.
The opening. In Spanish I open with personal experience because my primary audience is LATAM professionals who value the journey. In English I open with the data or the finding because the English-speaking audience wants to know first why it should matter to them. In Japanese I open with technical context because the Japanese technical audience values precision over personal narrative.
The length. Spanish tends to be longer. The grammatical construction of Spanish uses more words for the same idea. I do not force it to be shorter because naturalness matters more than length parity. Japanese is more compact by nature and can cover the same concept in fewer characters.
The examples. If I mention a bank, in Spanish I might reference the Colombian regulatory context. In English the global fintech context. They do not always change, but when cultural context matters, each version uses the most relevant one.
The slugs. Spanish has slugs in Spanish. English has slugs in English. Japanese uses English slugs because URLs with Japanese characters are problematic for sharing and for SEO.
The rules that do not change
Three rules apply to all three languages without exception.
No contractions in the blog. "Do not" instead of "don't". This applies in Spanish, English, and Japanese. Contractions are a tell of AI-generated content that was not reviewed.
No dashes. Never em dash, never en dash. In all three languages. Where others would put a dash, I use a period and a new sentence. Dashes are another tell of AI content.
Specific data always. "Gemini gave 9 out of 10, ChatGPT gave 8.2, Perplexity gave 7.5" instead of "scores varied." In all three languages. Data does not get adapted. Data is data.
Why it matters for GEO
The JSON-LD on each post connects the three versions through the workTranslation property. It tells AI models: this Spanish post, this English post, and this Japanese post are the same content in three languages. My hypothesis is that this consolidates authority instead of fragmenting it.
When an AI model searches for a source about GEO in Japanese, it can reach my Japanese version and attribute the authority of the English post (which has the freeCodeCamp backlink, the 602 prompts data, and citations from other posts). The three versions strengthen each other.
If they were literal translations, models might treat them as duplicate content in different languages. Since they are different versions of the same topic, each one contributes unique value. The signals I have observed suggest models treat them as complementary content, though I do not have definitive data on how each model internally processes connected translations.
In the llms.txt audit, all four AIs noted the multilingual structure as a differentiator. Claude described it as part of the intersection nobody else occupies: "designer + frontend builder + GEO practitioner + multilingual." The multilingual part is not an extra. It is part of the site's value proposition.
What most multilingual blogs get wrong
Most use automatic translation plugins. Content is translated word for word, published in 5 or 10 languages, and nobody verifies that the translation reads naturally. A native reader immediately detects that the text was translated because the grammatical structures are from the source language, not the target language.
Other blogs publish in one language and let AI models translate on the fly when someone asks in another language. That works for the model's response, but it does not build authority in that language. The blog has no presence in Japanese. The model only has the English version and translates internally.
Publishing adapted versions in three languages is more work. Each post takes more time. But each version builds presence and authority in its own language instead of depending on a model to translate on the fly.
The real cost
Generating the three versions with Claude takes between 15 and 30 additional minutes per post. Review adds another 15 to 20 minutes. Total: less than one extra hour per post to triple the linguistic reach.
Before AI, publishing in three languages would have required three writers or a professional translator per language. The cost would have been prohibitive for a personal blog. With AI as a writing partner, the cost is one hour of my time per post. The barrier that made multilingual content unviable for independent creators no longer exists.
The barrier that remains is the willingness to treat each language as its own text instead of as a translation. No tool solves that. Respect for the reader who will read in that language solves it.