Wouldn't it be better to program control of render to the front end? Similar to how we program the dark mode button? Sure, sending machines to consume a browser render is silly, but perhaps users would like to pile on the layers at will, MD/text/html/rich/full/etc... who knows what else is coming... Maybe "converse with site"
The problem is that only the human visitor may use JavaScript.
Content of my website (although I was lazy enough to not fill all sections) already markdown, rendered dynamically as html on client side.
But I rather keep LLMs from scraping it...
For other use cases like documentation, again the source (reStructuredText or Markdown) shall be consumed directly instead.
Why do you want to make it easier for AI agents to subsume your work?
Because you want the zero click Internet to be The Internet.
If you think about it, it's not very different from an RSS feed. And we don't consider the internet to be dead.
The article explains that saving tokens is an SEO technique.