An honest preface
Generative Engine Optimization is a real, measurable discipline — and a young one. In Indonesia, it is younger still. Anyone publishing dramatic "AI-citation lift" charts after eight weeks of work is either (a) cherry-picking cherry-picked prompts, or (b) making them up. We will not do either.
What follows is a description of the kinds of programs we run, the deliverables they produce, and the honest answer to "what's measurable right now." Specific client names are released only with consent; named, fully-detailed case studies are shared under NDA during a consultation.
The four-pillar program shape
Every GEO engagement we run organizes around the same four pillars. The mix changes by category and starting position.
1. Entity authority
Wikidata item built or completed, Knowledge Panel claimed, Crunchbase / LinkedIn / Google Business Profile aligned, schema.org Organization markup deployed on the brand's own site, and consistent NAP (name, address, phone) across the public web.
What's measurable: entity completeness score, presence of a Knowledge Panel, alignment of facts across sources. These are structural wins — they land within weeks and stay landed.
2. Content structure
FAQ schema on key pages, HowTo schema on procedural content, Article schema on long-form, definition lists on educational content, llms.txt and llms-full.txt published, canonical tags audited. The goal is to make the brand's existing content extractable — not to write more of it.
What's measurable: schema coverage by template, validation status (Google Rich Results, schema.org validator), and crawl-time AI bot accessibility (GPTBot, PerplexityBot, Google-Extended, Bingbot allowed and reaching content).
3. Citation outreach
Brand mentions and links from sources AI engines weight heavily for the brand's category and geography. For Indonesian brands this typically means industry publications (Tech in Asia ID, DailySocial, sectoral trade press), credible directories (Clutch, DesignRush, Sortlist, GoodFirms, Crunchbase), Wikipedia where notability is justifiable, and partner / customer pages.
What's measurable: referring domains added, quality (independent authority score), and the rate at which AI engines surface those new sources when asked about the brand or its category.
4. Multi-engine tracking
A curated prompt set (typically 60–150 representative questions) run weekly across ChatGPT, Perplexity, Gemini, Claude, Bing Copilot, and Google AI Overview, from Indonesian IP addresses, in both English and Bahasa Indonesia, with consensus runs to filter noise.
What's measurable: citation share per engine, share-of-voice vs named competitors, prompt-level wins and losses, and the trend of those numbers over time.
Three program patterns
Pattern A — Established brand, weak entity surface
A well-known consumer brand with strong organic SEO but a thin Wikidata item, no Knowledge Panel, and inconsistent founder / product information across the web. This is the most common starting state for Indonesian heritage brands.
Typical first 90 days: Wikidata item built and accepted, Organization schema deployed sitewide, FAQ schema added to top 20 pages by traffic, llms.txt published, prompt-set baseline established across six engines.
What we'll claim: structural completeness, baseline measurement now in place, leading indicators tracked.
What we won't claim: a specific revenue lift attributable to AI engines in the first 90 days. The honest answer is that we do not yet have the evidence base to make that claim defensibly in the Indonesian market for any brand at this stage.
Pattern B — Challenger brand, strong content, low entity
A challenger in a competitive category — fintech, healthtech, D2C — with strong product content but limited third-party citations and no entity work done. AI engines describe the category but rarely cite this brand.
Typical first 90 days: entity foundation built, top-of-funnel content rebuilt for AI extractability, citation outreach scoped to 8–12 high-priority Indonesian sources, weekly tracking live.
Leading indicator we watch: citation share movement on the 20 most commercially valuable prompts. We share these movements honestly — including the prompts where the brand is still losing.
Pattern C — New entrant defining a category
A brand entering or naming a category where AI engines have no clear default to recommend yet. The opportunity is to become the default before competitors define the conversation.
Typical first 90 days: definitional content built (glossary, "what is X" canonical pages, comparison pages), HowTo and DefinedTermSet schema deployed, citation outreach focused on category-defining publications, prompt set built around category-level questions rather than brand-name queries.
What's defensible: share of citations on category-level prompts, growth of brand-name prompts that previously returned "no result". We track both, weekly.
What we will publish — and when
As programs accumulate enough longitudinal data to be defensible (consensus runs, multi-engine, multi-month, Indonesian IP, against pre-registered prompt sets), we will publish named case studies with the underlying methodology disclosed. We would rather publish three honest cases than thirty inflated ones.
Want named, NDA-detail references?
We share named GEO references — including which engines they're winning, which they're not, and what the prompt-level data looks like — under NDA in a consultation call. If you want to see the real picture, that's the conversation to have.