I Used Codex to Build, Structure, Validate, and Deploy an SEO Glossary

I recently ran a small experiment with Codex that changed how I think about website production.

The goal was not just to generate content.

The goal was to see whether AI could help operate the website production workflow itself.

In this experiment, I used Codex to build an SEO glossary for Time Columns, an owned media site operated by Time LLC.

The first version included:

• 572 terms


• 26 glossary pages


• category classification


• short definitions


• HTML generation


• internal link checks


• sitemap updates


• GitHub push


• Cloudflare Pages deployment

From implementation to validation and deployment, it took about 95 minutes.

That is not just faster writing.

That is a different production model.

The usual website production workflow

A typical website production workflow often starts with structure.

You prepare a spreadsheet.


You define the sitemap.


You decide CMS fields.


You organize drafts and categories.


Then you move everything into the website.

The usual flow looks something like this:

Spreadsheet

→ Sitemap

→ CMS structure

→ Draft management

→ HTML / CMS implementation

→ Validation

→ Deployment

This approach is still useful, especially for large projects.

But it also creates many handoffs.

Planning moves to writing.


Writing moves to design.


Design moves to coding.


Coding moves to CMS entry.


CMS entry moves to checking.


Checking moves to deployment.

Each step may be simple, but every handoff adds time.

What I tried instead

This time, I tried the opposite direction.

Instead of preparing the perfect spreadsheet first, I started with the website.

I provided Codex with a list of terms and the purpose of the glossary.

The purpose was not to create a dictionary.

The purpose was to create SEO entry points inside the site.

From there, Codex helped with:

• grouping terms into categories


• removing duplicates


• writing concise definitions


• generating glossary pages


• updating related links


• checking internal links


• updating sitemap.xml
  


• preparing changes for GitHub


• deploying through Cloudflare Pages

The workflow became closer to this:

Term list

→ AI categorization

→ HTML pages

→ Link validation

→ Sitemap update

→ GitHub push

→ Cloudflare Pages deployment

→ Structured data extraction afterward

The interesting part is that the structured data could also be recreated afterward from the published HTML.

So the direction was reversed.

The website came first.


The structured data came later.

This was not just AI writing content

Most discussions about AI and websites still focus on writing.

Can AI write blog posts?


Can AI generate landing page copy?


Can AI rewrite SEO articles?

Those are useful questions, but they are not the most interesting part.

In this experiment, AI was not only writing text.

It was helping connect writing, structure, implementation, validation, and deployment into one continuous workflow.

That is where the leverage appears.

The human role still mattered

This was not “AI decides everything.”

The first attempt to let AI choose all the terms by itself produced something too thin and scattered.

The useful version started when I provided the term list and the direction.

The human role was still essential:

• deciding the purpose


• choosing the theme


• defining what the glossary was for


• judging whether the categories made sense


• correcting the structure


• deciding when it was good enough to publish

AI handled execution.

But the concept came from the human side.

That distinction matters.

AI can move fast, but it needs direction.

Why this matters for web development

For content-heavy websites, the bottleneck is often not the HTML itself.

The bottleneck is the workflow around it.

• Who prepares the content?


• Who defines the structure?


• Who updates the CMS?


• Who checks the links?


• Who updates the sitemap?


• Who deploys?


• Who fixes small issues afterward?

AI can reduce the friction between these steps.

It does not remove the need for planning.

But it changes how much planning has to happen before anything can be published.

For owned media, FAQ sites, glossaries, documentation sites, and SEO content structures, this is important.

A website can be built, validated, reorganized, and expanded much more quickly than before.

The bigger shift

The real change is not that AI makes web pages faster.

The real change is that AI lowers the cost of restructuring.

That changes how we think about websites.

A site no longer has to be perfectly designed as a database before publishing.

It can be grown first, then reorganized as the structure becomes clearer.

AI makes that kind of operation much easier.

Conclusion

AI is often described as a writing assistant.

But that description is too small.

In this experiment, Codex helped build pages, organize categories, check internal links, update the sitemap, push to GitHub, and deploy through Cloudflare Pages.

The website production workflow did not simply become faster.

It started to move in the opposite direction.

Instead of preparing all structure first and publishing later, we can now publish, structure, validate, and reorganize much more fluidly.

Original article:


https://column.time7.jp/en/column/ai-reverses-website-production-process/

Time Glossary:


https://column.time7.jp/en/glossary/