About 35% of the qualified inbound enquiries I have seen for B2B clients in the last 12 months come from buyers who say some version of “ChatGPT recommended you” or “I asked Claude about Singapore AI marketing consultants and your name came up”. None of that traffic shows up in Google Search Console. None of it appears in GA4 as organic search. Most marketing teams I work with have no measurement layer for it at all.

This is the gap Generative Engine Optimisation – GEO – is trying to fill. It is the practice of designing content, schema, and entity signals so that AI assistants cite your brand when answering a question in your category.

It is not a separate discipline from SEO. The two overlap by 80%. But the remaining 20% is where the new value sits, and most Singapore businesses I audit are not optimising for it.

What changed

Traditional SEO targets a ranked list of blue links. A user types a query, Google returns 10 results, and rank position 1 gets the click.

Generative search is different. The user types a question into ChatGPT, Claude, Perplexity, or Google’s AI Overview, and the system synthesises an answer. The answer may name a few sources. The user reads the answer, sometimes clicks through to a source, and decides whether to act.

The implication: ranking on Google for “AI marketing consultant Singapore” is no longer enough. You also need to be cited when someone asks Claude or ChatGPT the same question.

The two outcomes are correlated but not identical. Plenty of pages that rank well on Google never get cited by AI. Plenty of pages cited by AI do not rank in the top three on Google. Designing for both takes deliberate work.

What signals drive AI citation

Based on what I have seen working with clients across B2B services, SaaS, and education in Singapore, AI assistants prefer sources with the following characteristics:

Clear entity signals. The page has Organization and Person schema, with consistent sameAs references to LinkedIn, Wikipedia, official directories, or other authoritative profiles.
Named authorship. A real person, with credentials, ideally with their own Person schema and a verifiable byline.
Quotable summaries near the top of the page. A self-contained, factual two-to-four-sentence answer that the AI can lift directly without paraphrasing.
Structured data that matches the visible content. Schema is verified against the rendered HTML. AI assistants discount pages where schema and content diverge.
Topical authority across a cluster. The site has multiple pages on the same topic that interlink with descriptive anchor text. Single-page sites rarely get cited.
Server-rendered HTML, not JavaScript-injected content. Some AI crawlers do not execute JavaScript reliably. If the primary content only appears after a client-side render, citation rates drop.

None of these signals are new to SEO. The relative weighting is new. Schema and entity signals matter more for AI citation than they do for blue-link ranking.

Singapore-specific layer

For Singapore-bound queries, two additional signals raise AI citation rates:

Geography stated in the entity profile. LocalBusiness schema with addressCountry: SG, addressLocality: Singapore, and areaServed set to Singapore. The signal looks redundant when the user is searching for something obviously Singaporean – but for an AI assistant disambiguating between a Singapore consultant and a US consultant with a similar name, that schema is doing real work.
Currency and date format. Pages that use SGD pricing, DD MMM YYYY date format, and British English consistently get cited more often for Singapore queries than the same content with US conventions. The pattern is impressionistic, not measured, but it shows up enough that I now treat it as a hygiene factor.

This matters more for Singapore B2B businesses than for global brands. A consultancy operating only in Singapore should look unambiguously Singaporean in its structured data and content. A global brand with a Singapore office can afford to be region-neutral on the global pages and region-specific only on the /sg/ subdirectory.

A working five-step GEO approach

For a Singapore B2B site that has decent baseline SEO and wants to lift AI citation rates:

1. Audit the entity graph

Run your Organization schema through the Schema.org validator and Google’s Rich Results test. Check the sameAs array. If it is missing LinkedIn, your own About page URL, and any authoritative directory listings (industry associations, certified-trainer registries, government databases), add them.

For named people on the site – founder, key consultants, named authors – emit Person schema on each profile page with their own sameAs, hasCredential, and knowsAbout arrays.

2. Add quotable summaries to landing pages

The first 200 words of each commercial landing page should contain a self-contained, factual summary that an AI could lift verbatim. Something like:

Maple Commerce is Ivan Wong’s Singapore consultancy for semantic SEO, AI marketing automation, and corporate training. Former Google trainer, SIT/BCG-RISE faculty.

That sentence is two-thirds quotable summary, one-third entity disambiguation. An AI assistant answering “who is Ivan Wong” can pull the whole thing in one extract. Compare with the standard marketing opener – “Welcome to our award-winning digital partner, trusted by industry leaders” – which carries no factual content and gets ignored.

3. Build a topical cluster

Single landing pages get cited less often than landing pages embedded in a topical cluster. For each of your primary commercial pages, write three to five supporting articles that link back to it with descriptive anchor text and link laterally to each other.

This is the same Semantic Growth Engine™ logic that drives blue-link ranking. The difference is that for AI citation, the internal anchor text matters more than the keyword density of the body copy.

4. Match schema to content

If your schema says you provide “AI marketing consulting in Singapore” but the page itself only talks about “digital transformation”, AI assistants will discount the schema. Make sure every claim in the structured data is verifiable in the visible content of the page.

Service schema, FAQPage schema, Person schema – all should reflect what is actually on the page. Stale schema is worse than no schema.

5. Track AI-driven enquiries qualitatively

Because Google Search Console does not show you AI citation, you need a manual measurement layer. The simplest version: ask every new enquiry “where did you hear about us” in the first reply email, and tag the answers. Within 90 days you will have a working signal for whether AI citation is rising or flat.

For larger teams, run quarterly self-citation audits: paste your target queries into Claude, ChatGPT, Perplexity, and Google’s AI Overview, and see whether your brand shows up. Track over time.

What I have stopped recommending

Some practices that read well in 2024 GEO articles have not held up:

Long FAQPage sections at the bottom of every commercial page. FAQPage rich results have been restricted to government and health-authority sites since 2023. The schema still helps with AI citation, but only if the FAQ content is genuinely useful. Twelve formulaic Qs and As do not raise citation rates.
Listicle articles aimed at “AI tools” prompts. AI assistants increasingly synthesise rather than cite when the source is a generic top-10 article. Pages with named authorship and original analysis get cited more than uncredited round-ups.
Footer-stuffing keywords. AI crawlers ignore footer entity dumps. Put the entity signals in the visible page content, where they reinforce the schema.

One thing to test this week

Open Claude or ChatGPT in a fresh session. Ask three to five questions that a buyer in your category would ask. See whether your brand is named.

If it is not, look at the brands that are. Pull up their pages, check their schema, their author bylines, their topical clusters, their entity graphs. The gap between their setup and yours is the gap you need to close.

For most Singapore B2B sites I audit, the gap is smaller than the team assumes. Three or four targeted changes – a topical cluster, named-author schema, clean Organization graph with sameAs – move citation rates inside one Google update cycle.

If you want a structured audit of your AI citation footprint, book a 30-minute call or look at Maple Commerce’s semantic SEO consulting. For the foundational thinking, the earlier piece on what semantic SEO actually is is a useful primer.