Gen AI University

← Video Library · Case Study

How We Rebuilt This Website Agentically: Webflow to an Owned, AI-Native Site

Published July 23, 2026

Key takeaways

The website you are reading is the case study. We rebuilt genaiuniversity.com from the ground up, and we did it agentically, the same way we teach founders to run their companies. Here is exactly how it happened, and how you could do it to your own site.

The before: a rented site sitting on a goldmine

The old site was a page builder. It looked fine. But it was rented, not owned, and it sat on top of years of work that was not working as a system. Hundreds of videos lived only on YouTube, invisible to search on our own domain. There was no owned data. Several domains all served the same page, quietly splitting search authority instead of compounding it. The knowledge existed. It just was not organized into anything that could grow on its own.

What “own the site” actually means

Moving off a page builder onto an owned, HTML-first codebase is not a vanity move. A rented builder caps how far you can automate, because every change runs through a human clicking a UI. An owned site is something AI can operate on directly: read it, generate new pages, and deploy them without a person in the loop. That is the difference between a website you edit and a website that builds and grows itself.

The stack: built for crawlers and answer engines

We chose static HTML with zero client-side JavaScript, generated by Astro and hosted on Vercel with automatic deploys on every push. The reason is simple: search crawlers and AI answer engines read plain HTML best. Every page also carries structured data (video schema, breadcrumbs, organization markup) so both Google and AI systems understand what each page is and can cite it.

Turning 400+ videos into a searchable library, agentically

The core move was to give every long-form video its own page, with the full transcript, key takeaways, and question-style headers. That is answer engine optimization: a page that can be retrieved, cited, and that lands the viewer on our owned site with the video embedded, instead of only on YouTube.

The interesting part is how fast it went. Instead of writing 400+ pages by hand, we ran a swarm of agents, one per video. Each agent pulled its own transcript, wrote its own page, and did it in parallel with all the others. The entire library was built in a fraction of the time a person would need, at zero caption cost. (What an AI Chief of Staff is and how the same specialization principle works)

Keeping every ranking: the migration nobody notices

The real risk in any replatform is losing your search rankings and the backlinks pointing at your old URLs. We avoided it two ways. First, every old article kept its exact slug, so the URLs did not change. Second, a redirect map of more than 300 rules caught everything else: old campaign links, affiliate links, retired pages. The result is a migration that is invisible to Google. Rankings and links carry straight over.

The part that almost broke: one bad redirect

Here is the teaching moment. A single malformed redirect in the configuration quietly failed every deploy for a stretch, while the code itself was perfect. The lesson we drill with founders applies to your own build: validate configuration as ruthlessly as you validate content. The smallest error is the one that hides in plain sight, and it is almost always in the plumbing, not the product.

Going live: DNS, then the engines

Because everything was staged in advance, the cutover itself was pure DNS. Point the domain, make it public, done. Then we told the search and AI engines about it: submitted the sitemap to Google Search Console and Bing Webmaster Tools, and pushed every URL to IndexNow, which feeds Bing, Yandex, and AI answer engines almost instantly.

What it unlocks

Owned data and a site that grows itself. A weekly routine now detects new uploads and turns each one into a page automatically. The library expands without anyone touching it, and every new video becomes a new indexable asset the day it publishes.

This is what becoming agentic looks like

We did not use AI to help build a website. We built a system that builds and grows the website. That is the shift: from using AI to operating with it. If you run a real business with content and a domain, the same pattern is available to you. Own the surface, turn your existing assets into systems, and let AI run them alongside you.

The playbook, in six moves

  1. Extract the brand from the existing site, or write a creative brief if it is brand new.
  2. Scaffold an owned, HTML-first site built for crawlers and answer engines.
  3. Turn every asset into an indexable, schema-rich page, agentically, in parallel.
  4. Preserve every URL with exact-match slugs plus a redirect map.
  5. Instrument it with analytics, structured data, and a sitemap.
  6. Cut over DNS and submit to the engines (Google, Bing, IndexNow).

References

Want to build agentic systems like this into your own business, not just your website? That is the whole point of Gen AI University. Become agentic.