The Full Circle: Why the Old Rules of the Web Are the New Rules Again

How three decades of web evolution brought technical craftsmanship back from the archive

There’s a particular kind of satisfaction that comes from being right about something you didn’t even know you were predicting. For those of us who learned to build websites in the 1990s and early 2000s — hand-coding HTML, wrestling with CSS floats, obsessing over file sizes and load times — the last decade felt like a long, slow obsolescence. Our hard-won expertise seemed quaint in a world of drag-and-drop builders and bloated WordPress themes.

But something has shifted. Quietly, then suddenly, the web has come full circle. And the skills that were supposedly archived are now, once again, the edge that separates the practitioners from the pretenders.

When Constraint Was a Teacher

Building websites in the early web era was an act of disciplined engineering. Dial-up connections and primitive browsers punished carelessness. Every kilobyte mattered. Semantic HTML wasn’t a philosophical preference — it was the only way to make a page behave reliably across the patchwork of browsers and operating systems that existed at the time.

You had to understand what you were doing. There was no safety net.

Those constraints bred a generation of developers who thought deeply about structure, performance, and meaning. We didn’t just ask does this look right? We asked why does this work, and what breaks if I change it? That kind of thinking leaves a mark.

The Great Democratisation — and Its Hidden Cost

Then came the platforms. WordPress themes, and later Wix, Squarespace, and a dozen others, did something genuinely valuable: they opened the web to people who had ideas but not necessarily the technical background to execute them. Designers, marketers, and advertising agencies could suddenly ship beautiful, functional websites without writing a line of code.

This was not a bad thing. Access matters. The web became richer and more diverse because of it.

But something was lost in the translation. The people building these sites were optimising for the visual — understandably so, because that’s what clients see and approve. The code underneath was a different story: bloated templates, unoptimised images, render-blocking scripts, mountains of unused CSS, and not a schema tag in sight. Websites looked polished on the surface and were a disaster beneath it.

For a while, that didn’t matter. Google’s algorithm was forgiving enough, broadband was fast enough, and clients were happy enough that performance debt stayed hidden. The specialist web developer became a rarer creature, squeezed out by agencies who could now handle front-end work in-house.

The Reckoning

The reckoning arrived in stages.

First, Google started making noise about page speed. Then Core Web Vitals became a ranking signal. Then structured data — schema markup — became the language through which search engines and, increasingly, AI systems make sense of content. And then the most significant shift of all: the rise of answer engines.

We are no longer building websites purely for human eyes browsing a results page. We are building for crawlers, for large language models, for AI systems that synthesise information from across the web and surface it in direct answers. The question these systems ask is not does this page look professional? It is can I understand, trust, and efficiently parse what this page is saying?

The answer to that question lives in the code — not the design.

What the New Era Actually Demands

The skills that are suddenly valuable again are not, at their core, different from the ones that mattered in 1998. They have simply been recontextualised.

Semantic HTML — using the right element for the right purpose — matters now because AI systems and crawlers use document structure to infer meaning. A <article>, <nav>, <main>, and <h1> hierarchy tells a machine what kind of content it’s looking at. A <div> soup tells it nothing.

Schema markup — structured data in JSON-LD or Microdata — is the vocabulary through which a page declares its own identity to search engines and AI. Is this a product? A recipe? A business? An event? An article with a named author and a publication date? These declarations directly influence whether your content appears in rich results, knowledge panels, and AI-generated answers. Most drag-and-drop builders handle this poorly or not at all.

Performance engineering — minimising render-blocking resources, optimising the critical rendering path, managing Core Web Vitals like Largest Contentful Paint and Cumulative Layout Shift — is now a direct ranking factor and a determinant of whether an AI system considers your site a quality source. A slow site signals a careless site.

Content architecture — the logical, hierarchical organisation of information on a page — determines how well a crawler or AI can extract meaning. This is partly about HTML structure, partly about writing discipline, and partly about genuinely understanding how machines read.

None of these are skills you develop by dragging a widget into a template.

The Depth That Cannot Be Faked

Here is what years of building websites from scratch gives you that no platform abstracts away: an intuitive understanding of cause and effect.

When you have written the HTML by hand, you know what a heading hierarchy does. When you have debugged CSS specificity wars at midnight, you understand the cascade. When you have watched a site crawl in Google Search Console and traced the problem back to a missing canonical tag or a malformed schema object, you understand the relationship between code and discoverability in a way that is difficult to acquire from a YouTube tutorial.

The graphic designers building in Squarespace are not incompetent. The self-taught WordPress developers are not lazy. They are simply working at the wrong layer of abstraction for what the current web demands. They can see the interface. They cannot see — or meaningfully change — the infrastructure underneath.

That gap is the edge.

The Irony of the AI Era

There is a particular irony worth naming. The rise of AI is, for many people, a source of anxiety about obsolescence. Automation threatens to commoditise tasks that once required expertise. And yet, in the specific domain of web development, the AI era has done the opposite: it has raised the value of deep technical knowledge.

The more AI systems become the primary interface through which people discover information, the more important it becomes to speak the language those systems understand. And that language is not beautiful imagery or clever copywriting or a well-chosen WordPress theme. It is clean, semantic, structured, performant code — with schema that tells the machine exactly what it’s looking at.

The web has come full circle. The people who spent years learning to build it properly are not obsolete. They are, quietly and perhaps unexpectedly, indispensable again.

What This Means Going Forward

For those of us who carry the institutional memory of the early web, the moment calls for a kind of recalibration — not nostalgia, but translation. The principles are the same. The vocabulary has expanded. Schema is the new alt text. Core Web Vitals are the new file size limits. AI readability is the new cross-browser compatibility.

For agencies and marketers who built practices on the assumption that any designer could build a website: the audit is overdue. The gap between a site that looks good and a site that performs — in search, in AI visibility, in actual business outcomes — is widening. The people who can close that gap are not the ones who know which Squarespace template to choose.

For clients: the cheapest website is rarely the most expensive one to fix. Performance debt, invisible at launch, compounds over time in lower rankings, higher bounce rates, and diminished presence in AI-generated answers.

The web rewards craftsmanship again. It always did, if you knew where to look.

Written by someone who first hand-coded a website on a 56K modem and is, against all reasonable expectation, more relevant now than they were five years ago.