AI / Multilingual Site Operations

Autonomously Expanding a Japanese Website into an English Site

Using a Japanese site as the base and expanding an English section with AI

Time Columns expanded its Japanese column site into an English section. This was not just translation: it included English articles, an English top page, English privacy policy, About page, hreflang, sitemap updates, internal-link checks, GitHub push, and Cloudflare Pages deployment.

Overview

Time Columns expanded its Japanese column site into an English section.

But this was not simply a translation task.

Based on the Japanese articles, generative AI and Codex handled English article creation, English top-page creation, English privacy policy setup, English About-page setup, Japanese-English hreflang, sitemap updates, internal-link checks, GitHub push, and Cloudflare Pages deployment.

In other words, this was an experiment in semi-autonomously expanding an English site from a Japanese site as the base.

Why Build an English Section

Topics such as generative AI, Codex, Cloudflare Pages, SEO, LLO, and AI leverage are not limited to Japan.

Information moves very quickly in the English-speaking technology sphere, and experiments are happening globally.

That makes it meaningful to make practical Japanese knowledge readable in English as well.

The purpose was not to start full-scale overseas sales.

The main purpose was to make Japanese article assets indexable in English, increasing the site's overall information volume, topical coverage, and search entry points.

An English section also helps search engines and AI systems understand the broader range of themes Time Columns covers.

Traditional Multilingual Website Production

Traditionally, building a multilingual site requires many steps.

First, the Japanese article is translated. Then the translation is reviewed, shaped into a page, placed under a language-specific URL structure, connected to navigation, and checked for links.

If SEO is considered, canonical tags, hreflang, sitemap entries, meta descriptions, titles, internal links, and language-specific top pages also need to be adjusted.

Multilingual production is not just translation. It is translation, editing, HTML production, navigation design, SEO setup, and publishing combined.

When handled manually, the management cost grows quickly as the number of articles increases.

What We Did

Time Columns added an English section based on the existing Japanese site.

Instead of using a separate subdomain, the English section was placed under column.time7.jp/en/.

This allows the English section to share the same GitHub repository, Cloudflare Pages environment, CSS, sitemap, domain asset, and operation flow.

The first translated and rebuilt articles focused on generative AI, Codex, SEO, LLO, and AI leverage.

The English top page was initially created as an article list. Google Sites articles were not mixed in; only real files hosted on Cloudflare Pages were included.

This Is Not Just Translation

If the task were only to create English text, a translation tool could do part of the work.

But actual site operations require more than English prose.

Publishing an English page requires URLs, links, metadata, footers, related pages, privacy policy, About page, sitemap entries, and deployment.

In this case, after creating English articles, the workflow also built an English article-list top page, added an English privacy policy and About page, configured reciprocal hreflang, added English pages to the sitemap, checked internal links, pushed to GitHub, and deployed automatically through Cloudflare Pages.

At that point, this is not translation. It is multilingual site expansion.

The Japanese Site Becomes the Mothership

The important point is that the Japanese site became the source base.

The Japanese articles already contain Time LLC's practical viewpoint, operational philosophy, and decision criteria. The English section did not need to start from zero.

The Japanese articles were used as primary information, then restructured so that English-speaking readers could understand the concept.

This was not simple word-for-word replacement. Expressions and structures were adjusted so the meaning would land in English.

For example, the Japanese title about website production reversing direction became AI Is Reversing the Website Production Workflow in English. It was not a literal translation, but a concept rebuilt for English readers.

AI Cannot Decide the Concept

AI handled many parts of the workflow: English article creation, HTML conversion, link adjustment, hreflang, sitemap updates, GitHub push, and Cloudflare Pages deployment.

But the initial concept was not decided by AI.

Use the Japanese site as the base. Place the English section under /en/, not a separate subdomain. Exclude Google Sites articles and use only real files on Cloudflare Pages. Make the English top page an article list. Start with topics such as generative AI, Codex, SEO, LLO, and AI leverage.

Those decisions were made by the human side.

AI can turn those decisions into implementation. But humans decide what should be done, why it should be done, how far to go, and what not to do.

AI can propose concepts. But humans choose the concept, give it meaning, and take responsibility for executing it.

The Cost of Multilingual Expansion Changes

Multilingual sites used to be expensive to operate. Translation, review, page creation, SEO setup, publishing, and ongoing maintenance all add cost.

For small companies or individually operated owned media, English versions often stayed out of reach.

Generative AI and Codex change that premise.

The workflow can include not only translation, but also English article creation, HTML conversion, link adjustment, footer changes, sitemap updates, hreflang setup, display checks, and GitHub push.

This means the bottleneck of multilingual expansion moves from translation to operational design and judgment.

Which articles should be translated? What URL structure should be used? How should the English top page work? How far should the English section be prepared? How should the privacy policy and About page be handled?

Humans make those decisions, and AI handles much of the implementation. That division lowers the barrier to multilingual expansion.

Time LLC's View

Time LLC does not see generative AI only as a writing tool.

Writing articles, converting them to HTML, adjusting CSS, checking internal links, updating sitemaps, pushing to GitHub, publishing through Cloudflare Pages, and expanding a Japanese site into English are all part of website operations.

By placing AI inside those operations, the speed of site improvement increases.

The English section is one example of that approach.

The purpose is not simply to publish in English. The purpose is to make practical knowledge built in Japanese readable in another language and easier for search engines and AI systems to understand.

The English section is not just a translation layer. It is an operational asset that expands the site's information volume, topical coverage, and search entry points.

Summary

What we did was not simple translation.

Using the Japanese site as the base, we expanded English articles, an English top page, English privacy policy, About page, hreflang, sitemap, internal links, and deployment as one connected workflow.

Build practical knowledge in Japanese, make it readable in English, and lower the expansion cost with AI.

This flow will become important for owned-media operations going forward.