The AI SEO Playbook Has Changed. Here's What's New.

Traditional SEO rewarded whoever could dominate the ten blue links. AI search doesn’t care about your rank — it cares whether a machine trusts you enough to quote you. That’s a fundamentally different game, and most teams haven’t updated their playbook.

The Top-of-Funnel Is Being Quietly Dismantled

Moz’s Tom Capper put it plainly in a recent Whiteboard Friday: AI overviews and answer engines are collapsing the top-of-funnel as we knew it. Informational queries — the ones that used to feed brand awareness and content marketing pipelines — are increasingly resolved inside the chat interface. The user never clicks through. For Southeast Asian brands running content-heavy acquisition strategies, this isn’t a future concern; it’s a current one. Google’s AI Mode and Perplexity already handle a meaningful share of English-language research queries across the region. The implication is structural: volume metrics and keyword rankings are increasingly unreliable proxies for actual search visibility. A page can rank first and still be invisible to the growing slice of users who get their answer synthesised before they ever see a result.

Four Shifts That Actually Matter in the New AI SEO Playbook

Capper’s framework identifies four adaptation levers worth taking seriously. First, third-party citation building — this is the new link building, except the target isn’t PageRank, it’s LLM training data and retrieval sources. Getting quoted in authoritative publications, industry databases, and structured knowledge sources is now a direct input to AI answer quality. Second, structured entity clarity — AI models construct answers around entities, not documents. Your brand, products, and spokespeople need unambiguous, consistent representations across the web. Third, bottom-of-funnel content protection — while top-of-funnel erodes, comparison and decision-stage content still drives clicks. Doubling down on high-intent pages is defensible. Fourth, rethinking measurement — tracking impressions and AI citations alongside clicks gives a truer picture of visibility. Without this, teams are optimising blind.

---

The llms.txt Question: Signal or Noise?

Ahrefs recently released a free llms.txt generator built on analysis of 137,000 domains — and the tool itself is less interesting than the debate it surfaces. The llms.txt standard proposes a robots.txt-style file that tells AI crawlers how to interpret your content. In theory, it gives publishers a layer of control over how LLMs read and represent their site. In practice, Ahrefs’ own analysis suggests adoption is thin and the signal it sends to models is unproven. The honest position: implementing llms.txt costs almost nothing if you’re already maintaining a content structure, so it’s worth doing — but treating it as a GEO silver bullet is premature. For Southeast Asian brands operating multilingual sites across Thai, Bahasa, Vietnamese, and English, the more urgent priority is ensuring your core content is machine-readable, properly structured with schema, and consistently attributed across platforms like Shopee, Lazada, and LINE’s ecosystem. AI models scrape the open web; make sure what they find about you is accurate and coherent.

What X’s Bot War Teaches Us About AI Content Quality

X’s real-time public documentation of its 24-hour fight against chatbot spam — reported by Search Engine Journal — might seem tangential to SEO strategy, but it reveals something important about where AI content quality signals are heading. The spam X was combating was sophisticated enough that detection required continuous model updates throughout the day. The arms race between generative spam and platform detection is accelerating, and search engines are watching. Google has been explicit that AI-generated content isn’t penalised by default, but content that lacks genuine expertise, original perspective, or verifiable attribution increasingly gets filtered out of AI-cited sources. For brands in Southeast Asia tempted to scale content production purely through AI generation: the short-term output gain is real, but the long-term citation risk is significant. Machines are getting better at identifying content that was produced by machines for machines — and systematically discounting it. The brands that will be quoted by AI search in 2027 are the ones publishing content that demonstrably required a human to know something.

Key Takeaways

• Shift your KPI framework to track AI citations and brand mentions in LLM outputs alongside traditional click and ranking metrics — visibility has split into two parallel planes.
• Prioritise third-party citation building in authoritative regional publications and structured data sources; this is now a direct input to how AI models represent your brand.
• Treat llms.txt as a low-cost hygiene move, not a GEO strategy — structured schema, entity clarity, and original human expertise are the actual levers.

The uncomfortable question for marketing directors heading into Q3 planning: if half your top-of-funnel traffic is being answered before it reaches your site, what is your content strategy actually optimised for? The teams that will pull ahead aren’t the ones who find a new technical trick — they’re the ones who rebuild their content philosophy around being genuinely worth quoting.

---

At grzzly, we work with marketing teams across Southeast Asia who are rethinking their search visibility strategy for an AI-first environment — from GEO audit frameworks to citation-building programmes that work across English, Bahasa, and Thai. If your team is trying to make sense of what the new search landscape means for your brand specifically, Let’s talk