The search behaviour shift nobody's site was built for
For twenty years, being found online meant one thing: rank on page one of Google. That game hasn't ended, but a second one has started alongside it. A growing share of buying research now happens inside AI assistants — someone asks ChatGPT "what's a good content agency in Manila?" or sees a Google AI Overview summarise their question before a single blue link appears.
In those moments, the assistant composes an answer from sources it considers clear, credible, and quotable. If your site is one of those sources, you get named and linked. If it isn't, your competitor does — and the customer may never scroll far enough to discover you existed.
GEO (generative engine optimisation) and AEO (answer engine optimisation) are two names for essentially the same discipline: making your content the thing AI reaches for. Some practitioners use AEO for featured-snippet and voice-search style question answering, and GEO for LLM citation specifically. In practice, the tactics converge.
How AI assistants actually choose what to cite
AI assistants pull answers from two places: what they learned during training, and what they retrieve from the live web when browsing. You can influence both, but retrieval is the faster lever. When an assistant searches on a user's behalf, it behaves like a speed-reading researcher: it scans for pages that answer the exact question directly, checks whether the source seems legitimate, and lifts the clearest passage it can attribute.
That gives you a working checklist of what "citable" means:
- The answer appears early. A page that takes 600 words to reach the point loses to one that answers in the first paragraph and elaborates after.
- The structure is machine-legible. Descriptive headings, short sections, tables, and lists let the model locate the relevant part fast.
- The entity is consistent. Your business name, description, location, and offer read the same on your site, your Google Business Profile, your LinkedIn, and your directory listings. Conflicting data makes models hedge — and hedging models don't cite you.
- There's something only you could say. Original data, real client numbers, local specifics, first-hand experience. Generic re-explanations of what's already everywhere give the model no reason to prefer you.
GEO vs classic SEO: what carries over, what's new
| Classic SEO | GEO / AEO | |
|---|---|---|
| Goal | Rank as a link on results pages | Be quoted inside AI-generated answers |
| Unit of success | Position, clicks | Citations, mentions, assisted conversions |
| Writing style | Keyword-targeted, comprehensive | Answer-first, quotable, entity-clear |
| Key markup | Titles, meta, internal links | All of that + FAQ/Article/Organization schema |
| Crawlers to welcome | Googlebot, Bingbot | + GPTBot, ClaudeBot, PerplexityBot, Google-Extended |
| Time to first result | Often 3–6 months | Often 1–3 months via live retrieval |
The good news: nothing in GEO fights against SEO. A page built to be cited is also, almost by accident, a page built to rank. You're not choosing between the two games — you're playing both with one set of moves.
Nine practical steps to get cited
- Answer in the first two sentences. Put a complete, standalone answer directly under your H1, then expand. This is the single highest-leverage change.
- Phrase headings as the questions people ask. "How much does content marketing cost in the Philippines?" beats "Our Pricing Philosophy" every time.
- Add an FAQ section to key pages — with FAQPage schema. The Q&A format maps one-to-one onto how assistants receive questions.
- Mark up your organisation. Organization/LocalBusiness schema with your name, logo, location, and services gives models a machine-readable identity card.
- Unify your entity everywhere. Same name, same one-line description, same details on every profile and directory. Audit this quarterly.
- Publish something with actual substance. Local pricing data, process breakdowns, real results. One genuinely original page outperforms ten rewrites of what's already online.
- Welcome AI crawlers in robots.txt. Explicitly allow GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, and Google-Extended — being crawlable is the entry ticket.
- Keep pages fast and clean. Retrieval systems time out on slow, script-heavy pages. Static, semantic HTML is ideal.
- Show dates and authorship. A visible publish date, a named author or organisation, and updated content all feed the credibility signals models weigh.
A test we run on every page before it ships: read the first paragraph alone, out of context. If an AI assistant quoted only that paragraph, would it be a complete, accurate, useful answer that makes the reader want the rest? If not, it gets a third draft.
Why this matters extra in Singapore and the Philippines
Both markets are mobile-first, research-heavy, and increasingly conversational in how they search — and AI assistants handle conversational queries better than keyword search ever did. Meanwhile, most local SME websites still aren't structured for citation at all. That's a genuine early-mover window: the businesses that make themselves quotable now become the default answers the models keep reaching for, and defaults are sticky.
It's also a leveller. You don't need a huge domain or a decade of backlinks to be cited — you need the clearest, most specific, most structurally legible answer to the question. Small studios can win that on craft.