What ChatGPT recommends when a Singapore buyer asks.
We asked ChatGPT the questions a buyer asks, in English and in Chinese, and kept every answer word for word.
Pick a question. See who ChatGPT named.
The order is the order in the answer. Each question was asked once per language.
AskedWho does generative engine optimisation in Singapore?
Named in the answer
- 1Stridec
- 2First Page Digital
- 3AI Studio
- 4Hashmeta
- 5OOm
- 6PurpleClick
- 7MediaPlus Digital
- 8Canlah AI
The answer said it was compiled from published service pages and 2026 industry comparisons, and that independently verified client outcomes remain unclear.
All five answers as a list
- Who does generative engine optimisation in Singapore?Stridec, First Page Digital, AI Studio, Hashmeta, OOm, PurpleClick, MediaPlus Digital, Canlah AI
- Who helps B2B exporters in Southeast Asia become visible?Simaia, Geolix.ai, YCP Interactive
- Best platform for a small Singapore exporterAlibaba.com, Exporters.SG, LinkedIn, Global Sources
- CRM for a small B2B trading companyZoho, HubSpot, Pipedrive, Odoo, Dynamics 365
- How a manufacturer gets recommendedAlibaba.com, Global Sources
We ran GEO on ourselves first. Then AI started sending buyers.
TIY is our first client: a technical software house since 2013, selling JMP, Mastercam and Cast-Designer. Before selling this work to anyone, we did it on our own site, listings and records.
- Measure firstThe same buyer questions every month, on ChatGPT, Gemini, Perplexity and Google. Every answer kept word for word.
- Fix at the sourceWhere the company record, the directory listing and the site disagreed, we corrected the facts instead of writing new pages around them.
- Re-test and publishSame questions the next month, and the reading goes out whether it improved or not.
“I looked this up with AI. What we need is JMP software.”
- 12
- enquiries received
- 3
- said they found us through AI
- Aug–Oct 2026
- the period we counted
What this is not. We have no reading from before the work started, so this is sequence plus what the buyer said themselves, not proof of cause. TIY is our own company, not a paying client. Twelve enquiries is a small count, and three of them could move next month. How we counted.
Four steps, the same every month.
- 1
Fix the questions
The same set of buyer questions each time, so results can be compared.
- 2
Ask on the platform
Each question on ChatGPT, in English and in Chinese. Other platforms follow.
- 3
Keep every answer
Word for word, with date, platform and question. Nothing summarised away.
- 4
Count who is named
Counts first. Rates only after three rounds. Published every month.
Three things we learned, and what we do about each.
We did not start from a service list. We started by asking the same questions a buyer asks, and keeping what came back.
The shortlist is short, so we measure repeatedly.
One question named eight firms. A list that short can change with one new page or one new listing, so we run the same questions every month and track who enters and who leaves.
AI builds from published facts, so we fix the source.
The answer said it was compiled from published service pages and that verified client outcomes were unclear. So we start with the facts a machine can check: legal name, UEN, address, services, directory listings, and make them agree everywhere.
One answer proves nothing, so we publish counts.
Each question was asked once per language, which is enough to see who is named and not enough to claim a rate. We keep every answer, publish counts first, and only quote rates after three rounds.
Six questions to ask any GEO agency, and our answers.
You should be able to check us. These are the questions we would ask ourselves, and what we do about each.
We do not claim other agencies cannot do this. We claim you should be able to see it, and with us you can.
What this research does and does not show.
Shows
Who ChatGPT named on one day, for each question, and how it described them.
Does not show
A stable ranking, a market share, or how often a firm is named. One answer per question per language is a snapshot.
Next
A second reading in November 2026. We publish counts for three rounds before quoting any rate.
Ask any agency for what AI asks for.
The same answer that listed three agencies closed with a warning. It told the reader to demand evidence before hiring.
"Before hiring, ask each agency to test 20 to 30 real overseas procurement questions and show your current AI mention rate, competitor citations, and evidence of qualified buyer inquiries."
ChatGPT, answer to the third question, 25 October 2026That is the standard we set for ourselves, and the reason every answer on this page is kept rather than summarised. Take the requirement to any agency, including us. If they cannot produce the raw answers behind their numbers, the numbers are decoration.