Video: "Codex: New AI SEO Super Agent is INSANE!" by Julian Goldie on YouTube.
What Codex actually produced
Julian fed Codex a keyword and a short brief. What came back was not a draft in a text file — it was a complete HTML page: structured headings, body copy, meta title, meta description, schema markup, and internal link placeholders, all in one output. The page went from keyword to deployable file in under three minutes.
The key difference from a straight ChatGPT output is that Codex thinks in terms of code structure rather than prose. It approaches page-building the way a developer would: semantically correct HTML, logical heading hierarchy, and output that a CMS or static site generator can accept directly rather than needing reformatting first.
Why this is different from AI writing tools
Most AI writing tools produce a wall of formatted text. You still have to cut it into the right shape for your CMS, add schema, write the meta tags separately, and check the heading structure makes sense. That is not a huge amount of work but it adds friction at scale — across 100 pages it becomes a meaningful time sink.
Codex skips that step because it outputs the whole page as code. The content is already in the right containers. For sites that publish at any volume, that difference in output format matters quite a bit.
Where the output needs work
To be fair: the content itself is generic without careful prompting. Codex does not know your brand voice, your audience's specific questions, or the competitive context around your target keyword unless you tell it. What it does well is structure and format; what it does less well is the specific angle and insight that make a page worth ranking in the first place.
There is also the standard AI content caveat: factual claims need checking. Codex will write confidently about things it has inferred rather than knows. Any page going live needs a human read before it does.
The practical use case for a small business
The most realistic application is not replacing your content team — it is handling the mechanical parts of page production so the people who are good at writing can focus on the bits that require actual judgement. Use Codex to generate the structure and first draft, then have a writer or editor improve the angle, add real examples, and tighten the copy before it publishes.
For businesses with a long list of product or service pages to build — particularly in categories where the content structure is fairly consistent — this approach can cut production time substantially.
Where this connects to NordSys
Automated page production only pays off if the output is good enough to go live without hours of remediation. Getting there requires a solid prompt, a sensible workflow, and quality checks baked into the process. We build exactly these kinds of content pipelines for clients — from the initial Codex or Claude Code setup through to a repeatable process your team can run without a developer in the loop. Our Programming service covers AI-assisted content and build automation.
See our Programming service →