AI Search Optimisation: What Actually Changed and What Didn't

The pivot to AI search has produced a predictable wave of panic-driven content: new frameworks with new acronyms, breathless declarations that SEO is dead, and consultants rebranding themselves as GEO specialists overnight. Most of it overstates the rupture. Some of it misses the genuinely new risks entirely.

Here’s the cleaner read: the infrastructure of search hasn’t been replaced — it’s been extended. What has changed, quietly and consequentially, is where the failure points are.

The Fundamentals Didn’t Go Anywhere — But the Stakes Did

Ahrefs’ analysis of how to optimise for AI search lands on an uncomfortable truth for anyone selling disruption narratives: Google’s AI search features still run on its core search systems. Content still needs to be crawlable, indexable, and genuinely useful. Microsoft’s approach tracks similarly. The technical prerequisites for traditional SEO remain the prerequisites for AI visibility.

What shifts is the cost of getting it wrong. In a classic SERP, a thin page loses rankings but stays inert. Feed that same thin content into an AI-driven answer engine and it either gets ignored entirely or, worse, gets cited out of context in a summary that reaches thousands of users who never click through to verify. The margin for mediocrity has narrowed to near zero.

For teams in Southeast Asia managing multilingual content across Thai, Bahasa, Vietnamese, and English — often with inconsistent quality across language versions — this is the immediate operational risk worth stress-testing.

Bad Audience Data Is the Problem AI Will Make Visible

Search Engine Journal’s piece featuring Mallory Gray of Skydeo cuts closer to the bone than most AI-search commentary. The argument: AI models don’t fix poor audience signals, they amplify them. Mention counts — how often your brand appears across the web — are a weak proxy for actual purchase intent. What matters is behavioural signal quality: who is actually in-market, what they’ve done, and whether your targeting reflects that reality.

This has direct implications for how brands approach GEO (Generative Engine Optimisation). The instinct is to chase citations — get your brand mentioned in enough places that AI systems pick it up as authoritative. But citation volume without signal quality is the digital equivalent of buying reach without buying relevance. An AI agent optimising media spend on bad audience data won’t surface the problem; it’ll scale it.

For Shopee or Lazada sellers in the region running AI-assisted ad targeting, the practical implication is auditing your first-party data hygiene before trusting any automated optimisation layer sitting on top of it.

---

NLWeb and the Agentic Web: Structured Data Gets a New Job

Moz’s Crystal Carter offers the most forward-looking technical frame in recent weeks: NLWeb, built on the ASK protocol, is how your site talks to AI agents rather than just to crawlers. The Natural Language Web essentially treats your structured data as a conversation interface — agents query it directly to answer user requests without a traditional SERP in the loop.

The practical implication for SEO teams is that Schema markup is no longer a nice-to-have for rich snippets. It becomes load-bearing infrastructure for agentic visibility. A product page with clean, complete Schema is discoverable by an AI shopping agent. The same page without it is effectively invisible to that interaction layer — regardless of its organic ranking.

For brands operating across mobile-first markets like the Philippines or Indonesia, where a significant portion of commerce is already mediated through super-apps and chat interfaces (think GrabMart or LINE Shopping), the jump to agentic queries is shorter than it looks from a Western vantage point. The user behaviour is already there. The structured data often isn’t.

Implementation priority: audit existing Schema completeness for product, FAQ, organisation, and local business types. Then map which pages represent genuine commercial intent — those are your agentic surface area.

What ‘Optimising for AI Search’ Actually Requires Teams to Do

Collapsing the above into an honest operating picture: AI search optimisation is not a separate discipline. It’s traditional SEO with tighter tolerances, cleaner data requirements, and two new layers — structured data for agent readability, and audience signal quality for AI-assisted targeting.

The teams that will get this right aren’t the ones adopting the most tools. They’re the ones who can answer three questions cleanly: Is our content genuinely useful and technically sound at the page level? Is our structured data complete enough for an agent to act on? And does the audience data driving our AI-assisted campaigns reflect real behavioural signals or flattering proxies?

Semrush’s own framing of AI visibility improvement — while self-serving in context — at least acknowledges that search and AI visibility are converging metrics, not separate dashboards. That convergence is the right mental model. Treat them as one system with new failure modes, not two systems requiring parallel strategies.

---

Key Takeaways

• Structured data completeness is now the difference between agentic visibility and agentic invisibility — audit Schema coverage before chasing citation strategies.
• AI agents scale whatever data quality you give them: clean up audience signals before deploying AI-assisted targeting or media automation.
• For Southeast Asian brands managing multilingual content, the weakest language version of your site is now the ceiling on your AI search performance, not just your organic rankings.

The more interesting question isn’t whether to optimise for AI search — that’s table stakes now. It’s whether the organisations building these AI-assisted discovery systems are going to create more equitable visibility for locally-authoritative Southeast Asian content, or whether they’ll default to English-language citation patterns and call it global. The architecture is being written now. What gets baked into it matters.

---

At grzzly, we work with growth and digital teams across Southeast Asia on exactly this intersection — technical SEO foundations, structured data implementation, and the audience signal quality that determines whether AI-assisted strategies actually perform. If your team is trying to figure out where to focus first, let’s talk.