By Thanaphat Thinakul · CEO, IMAI Group
What does a GEO agency actually do for a company in Singapore? When a buyer asks ChatGPT "who are the best corporate secretarial firms in Singapore?" or asks Perplexity to compare HR software for a 50-person team, the assistant searches the web, reads a handful of pages and names a few companies. Generative Engine Optimization is the work of making sure your company is findable, understood correctly and mentioned in the places those assistants read, and then checking every week whether it is being named. This page explains how we do that, who we are, and the limits you should know about before you hire us or anyone else.
Who we are, plainly
ImVisible is built and operated by IMAI Group, an AI company in Bangkok, Thailand, led by CEO Thanaphat Thinakul. The same team builds the measurement platform and runs client programmes. Most of our clients so far are Thai businesses. We are now taking on companies in Singapore, Malaysia and the wider region, and we work remotely over email, LINE and video calls. Bangkok is one hour behind Singapore, so overlap with your working day is easy.
We do not have an office in Singapore, and we do not yet have Singapore client results to publish. We won't invent either. What we can show you before you sign anything is a free AI check-up on your own brand, with the actual answers the engines gave.
How AI assistants choose which companies to name
No engine publishes its selection logic, but the official documentation and independent research give a workable picture:
- Retrieval comes first. ChatGPT search relies on OpenAI's own crawler (OAI-SearchBot) together with third-party search providers, and Google's AI Overviews draw on the normal Google index. A page that isn't indexed, or that blocks these crawlers, can't be cited. Google states that there are no extra requirements to appear in AI features beyond being indexed and eligible for a snippet (Google Search Central; OpenAI crawlers).
- What others say about you matters more than links. Ahrefs studied 75,000 brands and found that branded web mentions correlated with AI Overview visibility at 0.664, against 0.218 for backlinks. A follow-up found YouTube mentions had the strongest correlation of all the factors it tested (Ahrefs; Ahrefs follow-up).
- Evidence-rich content gets used. The Princeton-led GEO paper (KDD 2024) found that adding statistics, quotations and cited sources raised a source's visibility in generative answers by up to about 40%, while keyword stuffing did not help (arXiv 2311.09735).
In our own Thai-market work, we have also found that pages hosted on the brand's own domain get cited, while client pages hosted on a shared agency domain almost never do. So everything we publish for you goes on your domain.
How we work, step by step
1. Baseline: a free AI check-up
AI answers vary from run to run, so a single screenshot proves little. Before any work starts we ask ChatGPT, Gemini, Perplexity and Claude a set of buyer questions for your category and market, three rounds over about five days. You get the answers as text, the competitors named instead of you, and the list of websites the engines cited. That last list tells us where your company needs to appear.
2. Question map
Buyers type short keywords into Google but ask assistants longer questions with conditions: "accounting firm in Singapore that handles Xero for startups", "best physiotherapy clinic near Tanjong Pagar that takes insurance". We build two lists, keywords to track in Google and questions to track in AI, and assign each one a single primary page so your own pages don't compete with each other.
3. Answer-first pages on your domain
Each page answers the question in its first paragraph, then gives the detail a buyer and a model both need: scope, process, typical timelines, price ranges if you publish them, limits, and an FAQ. Numbers appear only with a source. Package 1 includes 16 new articles a month; Package 2 includes 24 and adds English content for international buyers.
4. Consistent company facts
Your name, address, services, credentials and contact details should match across your website, Google Business Profile, LinkedIn, directories and review sites. When they conflict, assistants either skip you or get you wrong. We compile a fact sheet you approve, fix the about page and structured data, and list the places where the facts disagree.
5. Presence on the sites AI reads
Using the cited-sites list from step 1, we plan where you should appear: industry directories, comparison articles, review platforms, YouTube, and community threads where buyers ask for recommendations. We draft the listings and the copy. Posts on community platforms go out openly under your brand, by you or by us where the platform's rules allow. We don't write fake reviews, buy mentions or run sock-puppet accounts.
6. Weekly measurement
Every week we re-ask the engines, check Google AI Overviews for your keywords, and track your Google rankings and indexing. You see the % of questions where AI recommends you, per engine and per question, with the answer text saved as evidence. We judge progress by the trend across runs, not by one good day.
What you can expect, and when
| Period | Typical milestones |
|---|---|
| Weeks 1–2 | Baseline check-up · question map · fact sheet · blog set up on your domain |
| Months 1–2 | New pages indexed · first ranking movement · first listings live |
| Months 2–6 | AI starts recognising the brand on branded questions, then on "who should I use" questions as third-party mentions build up |
Unbranded "best X in Singapore" questions are the hardest, and low mention rates there are normal early on. We report branded and unbranded questions separately so you can see which is moving.
Our guarantee, and what nobody can guarantee
No agency controls what ChatGPT or Gemini says. If someone promises a guaranteed AI ranking, ask how they measure it. Our guarantee is narrow and written into the package: after 3 full months, if AI hasn't mentioned your brand at all in the questions we measure, month 4 is free of charge. The minimum term is 6 months.
Is this a fit for you?
Good fit: service businesses and B2B companies whose buyers research before contacting anyone, such as professional services, clinics, education, software and logistics. You should own your domain and be able to confirm the facts we publish.
Not a fit (yet): if you need leads this month, paid search is faster. If you'd want us to publish claims you can't support, we're the wrong team.
Pricing
We quote after the free check-up, because scope depends on your market, language and the number of keywords and questions. There are two packages (20 or 50 tracked keywords), described on our English homepage. For a deeper look at the two disciplines, read our AEO (Answer Engine Optimization) services and GEO vs SEO: what changes when buyers ask ChatGPT.
FAQ
Do you have a Singapore office or local staff?
No. The team is in Bangkok and works with regional clients remotely, in English.
Can you show Singapore case studies?
Not yet. We're new to the market and won't present Thai results as if they were local ones. The free check-up shows our measurement working on your brand.
Which AI engines do you measure?
ChatGPT, Gemini, Perplexity and Claude every week, plus Google AI Overviews for your tracked keywords.
Do we need llms.txt or special schema?
Not to be recommended. Google says no special markup is needed for AI features, and we haven't seen evidence that any engine uses llms.txt for selection. We add structured data because it keeps your company facts clear, not as a ranking trick.
How do we start?
Email contact@imaigroup.ai, message LINE @imaiofficial, call +66 98 789 3988, or run the free AI-visibility check.
Sources
- Google Search Central, "AI features and your website" — developers.google.com
- OpenAI, "Overview of OpenAI Crawlers" — platform.openai.com
- Ahrefs, "An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied)" and "Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews" — ahrefs.com
- Aggarwal et al., "GEO: Generative Engine Optimization", KDD 2024 — arxiv.org