HTML Should Be the Source Format for Digital Content
Slides, documents, videos, interactive explainers — the next generation of digital content should be authored in HTML. Not because it's trendy, but because it's the only format that is simultaneously human-readable, machine-writable, and universally renderable.
1. Better Expressiveness
For humans: beyond flat text
PowerPoint gave us slides. Word gave us documents. Each tool created its own island format, and each format carries the limitations of the era it was born in. HTML, by contrast, is a living standard — it has evolved over 30 years to support everything from simple text to embedded video, interactive charts, animated transitions, and responsive layouts that adapt to any screen.
When your source file is HTML, a slide deck can contain an embedded data visualization that updates in real time. A course module can include an interactive code playground. A research paper can embed a runnable experiment. No export step, no conversion loss, no "this feature is not supported in this format."
For agents: semantic, structured, and programmable
LLMs don't see pixels — they see text. HTML is the richest text-based format that still
carries explicit structure: headings, lists, tables, sections, and semantic
tags like <article>, <figure>, and
<nav>. An agent can read an HTML document and understand not just
what the content says, but how it is organized.
Compare this to a .pptx file: to an agent, it's a zip archive of XML blobs
with coordinate-based positioning. To modify a single bullet point, the agent must parse
OPC packaging, navigate slide1.xml, locate the right
<a:p> element among dozens, and reconstruct the XML without breaking
the slide layout. With HTML, the same change is a one-line text edit.
2. No Vendor Lock-in
The dominant content formats today — .pptx, .docx,
.key, .aep — are controlled by single vendors. They can change
the spec, discontinue the tool, or raise the price, and your content goes along for the
ride.
HTML is an open standard maintained by the W3C. Every browser renders it. Every operating system supports it. No single company can deprecate it. Your content, authored in HTML today, will be readable in 2040 without asking anyone's permission.
For the agent-human collaboration era, this matters even more:
- Agents can generate HTML natively. No need for specialized libraries or reverse-engineered file format specs. The agent writes text, and the text is the content.
- Humans can edit HTML in any tool. A browser, a text editor, an IDE, a visual editor — all of them work. There is no "you need this specific app to open this file" problem.
-
Version control is natural. HTML is plain text.
git diffshows you exactly what changed, line by line. Try diffing two.pptxfiles.
3. Composable with the Web Platform
HTML doesn't exist in isolation. It sits on top of CSS and JavaScript — the most widely deployed programming platform in history. This means an HTML source file can:
- Use CSS for visual design without embedding styles in the content structure.
- Use JavaScript for interactivity — animations, data binding, user input.
- Fetch live data from APIs and render it inline.
- Be embedded in other HTML pages, shared via URL, or printed to PDF.
No other content format offers this level of composability. A PPTX file is a dead artifact. An HTML file is a living application.
4. Accessibility by Default
HTML has built-in accessibility semantics. Screen readers understand
<h1> through <h6>, <nav>,
<table>, alt attributes on images, and ARIA roles.
When you author content in HTML, accessibility is a natural byproduct of good structure.
Proprietary formats require separate accessibility passes — and most authors skip them. The result: millions of PowerPoint presentations that are invisible to screen readers. HTML makes the accessible path the easy path.
5. The Agent-Native Content Layer
We are entering an era where AI agents will author, edit, and remix content alongside humans. This collaboration requires a format that both parties can read and write fluently. HTML is the only candidate:
| Capability | HTML | PPTX | Markdown | |
|---|---|---|---|---|
| Human-readable source | ✓ | ✗ | ✗ | ✓ |
| Agent-writable without libraries | ✓ | ✗ | ✗ | ✓ |
| Rich layout & design | ✓ | ✓ | ✓ | ✗ |
| Interactivity | ✓ | ~ | ✗ | ✗ |
| Native version control | ✓ | ✗ | ✗ | ✓ |
| Universal rendering | ✓ | ~ | ✓ | ~ |
| Embeddable in web pages | ✓ | ✗ | ✗ | ✓ |
Markdown is readable and writable, but it lacks layout, design, and interactivity. PPTX and PDF are visually rich, but opaque to agents and hostile to version control. HTML is the only format that scores well on every axis.
What Does This Look Like in Practice?
Slides
A slide deck as a single HTML file. Each slide is a <slide> tag.
An agent generates the content; a visual editor handles the layout. The same file is
viewable in any browser, editable in any text editor, and printable as PDF.
Interactive Tutorials
A coding tutorial with embedded <textarea> elements that run
JavaScript in real time. The agent writes the tutorial; the student interacts with
it — all in the same file format.
Research Papers
An academic paper with embedded charts that respond to hover, expandable methodology sections, and live data tables. No supplementary materials needed — everything is in the source file.
Video Storyboards
A video storyboard as HTML with timeline annotations, embedded preview frames, and cue notes. An agent generates the storyboard from a script; the director reviews and edits it in a browser.
The Missing Piece: Tooling
If HTML is such a great source format, why isn't everyone using it already? Because the tooling gap is real.
Writing raw HTML by hand is tedious. Existing visual editors (Figma, Canva, Google Slides) export to HTML but don't use it as their source format. Web IDEs understand HTML but aren't designed for content creation workflows — they think in terms of components and builds, not slides and chapters.
What we need is a new class of tools that:
- Use HTML as the native source format — not an export target.
- Provide visual editing for humans — drag, resize, style — without hiding the HTML.
- Expose a structured contract for agents — so AI can generate, verify, and edit content programmatically.
- Treat the source file as the single source of truth — no sidecar metadata, no hidden state.
This is what we're building at StarryKit.
Our first product, Starry Slides, uses a single HTML file as the source of truth for an entire slide deck. An agent can generate a deck from a contract, verify it against a schema, and open it in a visual editor — all without converting between formats. The human edits visually; the agent reads and writes the same HTML. No lock-in, no conversion loss, no gap between what the human sees and what the agent understands.
We believe this pattern — HTML as source, contract as schema, editor as interface — will extend beyond slides to documents, storyboards, dashboards, and formats we haven't imagined yet.
The Argument in One Sentence
HTML is the only content format that is simultaneously readable by humans, writable by agents, renderable by every browser, controllable by version systems, and extensible without anyone's permission. That's not a coincidence — it's the result of 30 years of open standards work. It's time to stop treating HTML as "just for web pages" and start treating it as the universal source format for digital content.