A Note on Last Year's Google Search Central Deep Dive

Last year I was invited to speak at Google Search Central Deep Dive APAC. My talk was titled The Best Lesson: Learned from a JavaScript SEO Disaster.

Looking back now, I want to write a small note about it.

What I talked about was JavaScript rendering — an old topic. Plenty of companies build beautiful sites with modern JS frameworks, and the experience is flawless for human visitors. But what Googlebot sees can be a blank page. This has been discussed in the SEO world for years, and yet new companies keep falling into the same trap.

I chose this topic because, after enough projects, I kept confirming the same thing: technical visibility is the "Step 0" of everything else.

Rankings, keywords, traffic, brand — all of these come later. Before any of them, there's a much more basic question: can the search engine actually read your content? If the answer is no, everything downstream collapses. You've already lost before the race begins.

What stuck with me from that event was how Google's own team framed the AI search question.

At the time, the entire industry was talking about AI Mode, AI Overviews, Gemini. Many people believed the rules of search had been rewritten. But Google's team kept making the same point: underneath all these new experiences, it's still the same flow — Crawling, Indexing, Serving.

I'd said this myself plenty of times. But hearing Google's own team confirm it lands differently. It means:

If your content isn't crawled, you won't appear in generative AI's answers. If your site isn't properly indexed, you won't be part of AI Overviews. If Google can't understand your products and services, you won't be served to your audience.

AI search didn't skip the index. It was built on top of it.

After that event, I became more confident in something I'd been saying: in the AI era, foundational SEO matters more than ever, not less.

Not because SEO hasn't changed. But because AI models aren't magic. They're learning machines that need data, and your website is the data. If your data is trapped behind broken JavaScript rendering, blocked by server misconfiguration, or structured so messily that crawlers can't parse it, AI has nothing to learn from.

In the world AI sees, your brand doesn't exist.

In the year since that talk, I've worked on more GEO projects and run more model-visibility tests. Most of what I've seen keeps confirming the same thing: what actually decides whether a site gets recommended by AI is rarely how flashy the content is — it's the more basic things, the Step 0 things. Whether it can be read. Whether it can be parsed. Whether it can be understood.

New tools will keep coming. AI will keep evolving. But as long as search engines and AI still need to see you before they can recommend you, technical visibility will always be the step that comes before everything else.

I don't expect this to sound like a breakthrough insight. But some things — the more basic they are, the easier they are to overlook.