3,900 AI-Generated Pages. 100/100 PageSpeed. 3 Weeks.

Written by, Adnan Khan on March 2, 2026

case-studyai-engineeringweb-developmenteducation

3,900 pages of professional course content.

Generated from old material using AI.

Deployed on a site that loads in under a second.

Total content creation time: a fraction of what it would take a team of writers working for months.

That’s not a typo. That’s what we built for CEOSH AB UK Ltd, an internationally accredited awarding body offering 70+ professional qualifications and roughly 5,000 CPD courses across health and safety, energy management, quality control, healthcare, fire safety, and ISO lead auditing.

This is the story of how we did it.

The Problem Every Content-Heavy Business Has

CEOSH had years of course content. Qualification specifications. Unit breakdowns. Learning outcomes. Assessment criteria. Entry requirements. CPD course descriptions across dozens of categories.

All of it existed. None of it was on a website that worked.

Most businesses with deep content libraries are in the same spot. You’ve got the material. It’s sitting in old PDFs, Word documents, spreadsheets, outdated CMS exports, or scattered across internal systems. You know you should get it online. You know it should be searchable, structured, and optimized for the people trying to find it. But every time you look at the scope of actually doing it, you close the laptop and move on with your day.

The math just doesn’t work.

Hire a content team to manually write, format, and publish 3,900 individual course pages, each with a title, description, category, level, duration, credit value, unit breakdowns, learning outcomes, entry requirements, and assessment methods. You’re looking at months of work and a serious budget.

So the content sits there. Gathering dust. Not working for you.

The AI Content Transformation

We took CEOSH’s existing course material, the raw content built up over years of developing accredited qualifications, and ran it through Claude to generate structured, publication-ready content at scale.

Every single piece of content on the CEOSH website was generated using AI. Not just the course pages. Everything. The ~3,900 individual course pages. The qualification descriptions for all 70+ programs. The website marketing copy. Page descriptions. CTAs. Recruitment funnel content for training centres, tutors, and assessors.

All of it. Generated from existing material. Structured for the web. Optimized for search engines.

In practice, it looked like this:

For each of the ~3,900 course pages, Claude generated structured markdown with typed frontmatter: title, description, category, level, duration, credits, units, learning outcomes, entry requirements, and assessment methods. Every page follows the same schema. Every page has proper SEO metadata.

For the 70+ qualification pages, Claude took existing qualification specifications and turned them into clear, scannable web content organized across 27 categories.

For the website copy, Claude generated the homepage value proposition, feature descriptions, stats sections, about page content, recruitment funnel copy, and CTA language. All aligned to CEOSH’s brand as an internationally accredited awarding body.

Old content in. Claude in the middle. Structured, web-ready pages out.

Why This Changes the Math

Think about the traditional approach.

You need 3,900 course pages. Each one needs to be written, formatted, and reviewed. Even if a content writer can produce 20 polished pages per day (that’s aggressive), that’s 195 working days. Nearly a full year of a single writer’s output. And that’s just the course pages. You’ve still got the website copy, the recruitment funnels, the category descriptions, the about page.

With AI-driven content generation from existing source material, we collapsed that timeline dramatically. The content was generated, structured, validated against typed schemas, and deployed, all within a 3-week project timeline that also included the entire website build.

If you’re a business owner sitting on years of old content, pay attention here.

You don’t need to hire a content team and spend six months on a migration project. You can repurpose what you already have. The raw material, the expertise, the course specifications, the product descriptions, the documentation. That’s the hard part, and you’ve already done it. Turning it into structured web content is a solved problem now.

The Technical Platform That Makes It Work

AI-generated content is only half the story.

The other half is what you do with 3,900 pages once you’ve got them. Throw 3,900 pages into WordPress with a MySQL database and every page load is a database query. Performance degrades. Server costs scale. Security surface area grows. You need caching layers, CDN configuration, plugin management, and regular updates just to keep things running.

We went a different direction.

Astro 5 Static Site Generation

Every one of those ~3,900 course pages is pre-rendered at build time into pure HTML. No server. No database. No runtime queries. When a user visits any course page, they get a static HTML file served from a CDN edge node.

Near-instant page loads. On every single page.

Content Collections with Typed Schemas

This is where AI-generated content meets engineering discipline. Every markdown file is validated against a typed schema using Astro’s content collections and TypeScript. If a course page is missing a required field (no title, no category, no description) the build fails.

The AI-generated content has a quality gate. Every page must conform to the schema or it doesn’t ship. Structure is enforced at the build level, not by someone manually checking 3,900 pages.

Zero Client-Side JavaScript

Astro ships zero JavaScript to the browser by default. The course pages, category listings, qualification browsers. All pure HTML and CSS. Interactive elements like the contact form and dark mode toggle use minimal, targeted scripts.

Nothing else.

That’s why the site scores 100/100 on Google PageSpeed Insights. No JavaScript bundle competing with content rendering.

Partytown for Analytics

Google Analytics and Tag Manager are necessary but notoriously heavy. We offloaded them to a web worker using Partytown. Analytics scripts run in a separate thread and never block page rendering or user interaction.

Self-Service Content Pipeline

This is the operational payoff. Adding a new course to the site doesn’t require a developer. Create a markdown file with the right frontmatter, drop it into the content directory, and the build system handles everything: page generation, category aggregation, navigation updates, sitemap inclusion.

CEOSH can add new qualifications and courses on their own. The system scales with their content, not against it.

The Numbers

Let me be specific about what was delivered.

What This Means for Your Business

If you’re sitting on years of content (course catalogs, product specifications, technical documentation, service descriptions, training materials, internal knowledge bases) this case study is a proof point.

The old way: hire writers, spend months, hope the formatting is consistent, manually publish each page, then maintain a CMS and database to serve it all.

The new way: take your existing content, use AI to transform it into structured, web-ready material at scale, deploy it on a static platform that scores perfect on performance metrics, and hand your team a simple content pipeline for ongoing updates.

3,900 pages. All AI-generated from existing material. All validated against typed schemas. All pre-rendered to static HTML. All scoring 100/100 on PageSpeed.

No WordPress. No database. No server to maintain. No content team working for months. And the whole thing runs on Vercel’s free tier. Zero hosting cost for 3,900 pages.

The Tech Stack

For the technically curious:

Ready to Transform Your Content?

If you’ve got existing content that should be working harder for your business (old course materials, product catalogs, documentation, training resources) we can help you turn it into a high-performance web presence.

No months-long content migration. No bloated CMS. Just your existing material, transformed and deployed on infrastructure that performs.

Start a conversation about what’s possible with your content.

elfmachines-ai

$ AI assistant coming soon...

In the meantime, start a conversation with our team.

> Type a message... _