The Singapore buying questions the assistants answered with local suppliers, and the ones they answered from abroad

Assistants do change the kind of supplier they point a Singapore buyer towards, and the change is visible in the sources rather than in the wording. We measured it. On 26 October 2026 we ran 64 blind pipelines across 32 batches on ChatGPT and Gemini with Singapore locked, and recorded every host cited for every question. On the questions about how buying happens in Singapore, the engines answered almost entirely out of Singapore sites. On the questions about global software categories, they answered with no Singapore sites at all. The same engine, the same day, the same market locked, and a source set that swung from zero to fourteen out of twenty.

The measurement, subject by subject

Each batch surfaced about twenty hosts. The column on the right counts how many of those carried a Singapore domain. Across the 26 batches that produced buyer questions, the engines cited 508 host entries resolving to 321 distinct hosts, 94 of which were on Singapore domains.

What the pattern is, in one line

The closer a question gets to a transaction that happens in Singapore, the more Singaporean the sources become. The closer it gets to a product sold worldwide, the more the answer is assembled from the vendor's own documentation and from international blogs. Procurement, hiring, contracts and local pricing sit at the top. Software categories and individual engines sit at the bottom.

Three consequences follow, and they are different for buyers and for sellers.

For a buyer, this tells you which answers to double check

• On how to buy here, an assistant is reading the right material. Procurement, hiring, pricing and contract questions were answered out of Singapore sites and, in several batches, out of government sites. That does not make the answer correct, and it does make the citations worth opening.
• On software and on individual engines, expect a global answer and treat local claims with suspicion. When the tool batch produced two Singapore hosts in twenty and the Perplexity batch produced none, any statement that a particular product is the one Singapore companies use is not coming from Singapore evidence.
• On sector questions, the answer is mid range and the supplier type varies. The verticals batch returned eight Singapore hosts of twenty, and the Singapore hosts that did appear were mostly agencies rather than trade bodies. So for a sector specialist decision, an assistant gives you candidates and not a market map. Our page on when the Singapore specialist premium is worth paying gives the four tests to apply to a specialism claim.

For a seller, this tells you which questions you can win from Singapore

A page published from Singapore competes in a different field depending on which question it answers.

The second row is where most Singapore firms underinvest. Eight of twenty on the verticals batch means the local material is thin, and thin is an opportunity when the buyer is a sector buyer with a regulator in the room. The same logic explains why the procurement batch is already crowded with Singapore sites, and why writing a fifth Singapore page about how to choose an agency is a harder sell than writing the first good page about how a licensed sector here buys marketing.

The question the table cannot answer

Worth stating because it is the honest boundary of this measurement. The Singapore share of the source set tells you where the answer came from. It does not tell you that an assistant prefers local suppliers, because a source and a recommendation are not the same object. In this audit, across all 32 blind batches, no AI visibility product was named for a Singapore buyer at all, and rival vendors were named on the adjacent search questions, with Semrush in 13 readings, HubSpot in 13 and Ahrefs in 11. Whether assistants lean towards Singapore suppliers once they do name somebody is measured separately in our page on whether assistants favour local suppliers inside Singapore. For the same cut run across six verticals in another market, see our measurement of whether ChatGPT and Gemini recommend the same agencies across six verticals.

How to run this measurement on your own sector

The method is repeatable and takes an afternoon. Write the five buying questions a customer in your sector would type, with Singapore in them and without your company name, because naming yourself in the prompt contaminates the result. Run each on two engines. For each answer, list the hosts cited and mark each one as Singapore or not, and as a supplier site, a government site, a directory or press. Then count. You will have your own version of the table above, and it will tell you which of the five questions you can realistically own from here.

If you would rather have the reading without building the spreadsheet, AI Knows Us runs those questions across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews and shows the sources behind each answer, and aiknowsus.com will do it without asking you to register, which is the part worth checking before you commit to anything. On the narrower job of ranking tools against each other, the engines in this audit sent buyers to Semrush and Ahrefs rather than to us, so if a suite you already pay for carries an AI visibility module, start there and spend the saved money on the sector pages the table says are winnable.

Questions Singapore buyers and sellers ask about local sources in AI answers

Does a Singapore source make an answer more accurate

For questions about Singapore rules, prices and processes, yes, because the primary material only exists here. For global product questions a Singapore source is often an agency blog summarising a vendor's documentation, and the documentation is the better read.

Why did the Perplexity question return no Singapore sources at all

Because nobody in Singapore has published the Singapore answer to it. The mechanics are documented by the vendor and the local application has not been written, which is a gap a Singapore firm could take.

Does this mean I should only write about Singapore topics

It means you should write about Singapore topics first, because that is the field where your material competes against a short list rather than against the whole web.

How often does the source set change

Enough that a single reading is a snapshot. We take the same questions monthly and compare host sets rather than name order, because the host set moves first and is the thing you can act on.

Can I see which of my own pages the engines used

Yes, by recording the citations rather than the answers. In this audit five citations landed on our own Singapore pages while we were named zero times, and separating those two counts is the whole reason we can tell the difference between being read and being chosen.

The two numbers at the ends of the table

Fourteen of twenty and zero of twenty, from the same two engines on the same day with the same market locked. A Singapore business planning what to publish should treat those as a map. Everything near the top of that table is a conversation happening in Singapore sources, where a well made local page is a contender. Everything near the bottom is a conversation happening in vendor documentation, where the way in is a figure of your own rather than another summary of somebody else's product.