SEO's always sat on four legs-content, links, tech, UX. That's not shifting. You still need those for search visibility and for generative engines, because they pull from the same SERP pool.
But after Google I/O, something clicked for me. We're moving toward a web that agents need to parse, not just people. That rewrites some of the rules under the hood. a lot of it folds back into UX and accessibility-can an agent navigate the visual layout, the HTML tree, the accessibility layer? If not, you're invisible to the machines that summarise for users. that's a new SEO discipline, not a separate one.
Then you've got the content layer: can an agent understand what your site is and does at a glance? That's still classic SEO, just reframed.
And then the fresh stuff: llms.txt, Markdown, agent-aware headers, content signals, WebMCP. Not for appearing in AI answers, but for enabling agentic action-your site being used as a tool, not just a source. That's where it gets properly technical.
If your definition of SEO includes making a site legible to autonomous agents, then yes, it's turning into more dev work. But links and content aren't going anywhere. They're just the foundation. The craft is adapting what sits on top.
