Getting a Singapore product recommended when a shopper asks an assistant what to buy, and the three catalogues that decide it

A product gets recommended by an assistant when it sits in one of the catalogues that assistant reads, with a price, a stock state and a product page that agrees with the feed. That is a feed and markup job, not a content job. Three routes publish their own terms and you can read all three today. Google's Merchant Center free listings put products into Search, Maps, Gemini, YouTube, the Shopping tab, Images and Lens at no media cost. Shopify sells an Agentic plan at "$0 SGD /mo" for selling in AI channels, which adds your products to Shopify's Catalog. OpenAI publishes the Agentic Commerce Protocol, an open standard for getting a catalogue into ChatGPT. Everything else a Singapore seller is sold on this subject sits on top of those three.

The three catalogues, with the page each quote came from and the day it was read

Each figure and each phrase below was read in the served page on 29 October 2026. None of it is promised to be current next month, which is why the date sits beside it.

Google Merchant Center free listings

Google states that with a free listing "the products you sell can appear at no cost on Google Search, Google Maps, Gemini, YouTube, the Shopping tab, Google Images, and Google Lens", read on support.google.com on 29 October 2026. The naming of Gemini in that sentence is the part worth stopping on. The same mechanism that has put a product into the Shopping tab for years is the mechanism that puts it in front of somebody asking Gemini what to buy, and it is the same feed, the same account and the same product data quality rules. Google also says free listings "help your product pages appear in search results, not other pages like homepages, shipping and return policies, or product category pages", so the unit of eligibility is the product page rather than the shop.

Shopify's Catalog and the plan that costs nothing

Read on shopify.com on 29 October 2026, the Singapore pricing page lists an "Agentic" plan at "$0 SGD /mo", described as "For selling in AI channels", with three lines under it: "Add your products to Shopify's Catalog", "Choose which AI channels to sell in" and "Only pay when you make a sale". Card rates on that plan are given as "from 3.2% + $0.50 SGD online". The paid ladder on the same page carries "Sell in AI chats" as an included feature on Basic, Grow, Advanced and Plus, with Basic at "$29 SGD/mo" paying yearly and "$39 SGD/mo" paying monthly, Grow at "$99 SGD/mo" and "$132 SGD/mo", Advanced at "$399 SGD/mo" and "$531 SGD/mo", and Plus "from $2,300 USD/mo". The third party transaction fee column runs from "2%" on Basic to "0.2%" on Plus. A Singapore seller on any plan therefore already has the AI channel entitlement, and the thing that decides whether it does anything is whether the catalogue has been submitted and whether the data is clean.

OpenAI's Agentic Commerce Protocol

Read on developers.openai.com on 29 October 2026, OpenAI describes it in its own words: "Agentic Commerce Protocol (ACP) is an open standard that serves as the connective layer between merchants and ChatGPT users. It enables ChatGPT to ingest structured catalog data, understand merchant inventory, and surface relevant products in context." The documentation set beside that sentence covers Feeds, Products and Promotions. So the shape of the ask is the same as Google's: a structured catalogue with inventory in it, submitted rather than crawled.

The robots file decides whether any of it is visible in a text answer

A catalogue handles the shopping surfaces. The ordinary written answer, the one a shopper gets when they ask which brand of humidifier to buy in Singapore, is assembled from pages a crawler was allowed to read. OpenAI's own crawler documentation, read on platform.openai.com on 29 October 2026, names four identities, GPTBot, OAI-SearchBot, ChatGPT-User and OAI-AdsBot, and says of the search one: "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links." It also says "it can take ~24 hours from a site's robots.txt update for our systems to adjust."

That makes a one line check the first thing to do before buying anything. Open your own robots file, look for a disallow against OAI-SearchBot, and if a developer added a blanket block during a scraping scare, you are paying for pages that are not eligible to be quoted. Our page on how to check a Singapore marketing supplier's own robots file before you hire it runs the same check on the agencies selling this work, and the results there are a fair guide to how often it is overlooked.

What has to be true on the product page itself

Google's product structured data documentation, read on developers.google.com on 29 October 2026, splits the markup into two classes and the distinction matters for a seller. "Merchant listings" are "For pages where customers can purchase products from you", with options for apparel sizing, shipping details and return policy information. "Product snippets" are "For product pages where people can't directly purchase the product", which is the editorial case. The page also states plainly that "Providing both structured data and a Merchant Center feed maximizes product visibility", so the two are not alternatives.

Four fields do most of the work, and all four are things a shopper would ask about anyway. Price, written the same in the feed and on the page. Availability, kept honest, because Google notes availability data is there "to help customers know when you have a product in stock". Shipping, including whether delivery is free, so the total is visible. Returns, including the policy and any fees involved. A catalogue with four accurate fields beats a catalogue with forty guessed ones, because a mismatch between feed and page is the most common reason a listing stops being eligible.

The Singapore layer, which changes the number a shopper is quoted

Price display is where a Singapore seller gets caught, because the figure in the feed is the figure an assistant repeats. IRAS publishes that "The current GST rate is 9%" and that a business "must register for GST when your taxable turnover exceeds $1 million", read on iras.gov.sg on 29 October 2026. If you are registered, the price a shopper sees has to be the price they pay, and a feed carrying a pre tax figure is a complaint waiting to arrive at the end of a checkout.

The second Singapore specific point is the currency of the fee rather than the price. Shopify's Singapore page quotes its plans and card rates in Singapore dollars and its Plus tier in US dollars, on the same page, read 29 October 2026. That is normal in this market and it is also the reason a seller comparing two quotes should ask which currency the invoice arrives in before comparing the numbers at all.

What two assistants actually did with this question

The instrument here is saved assistant answers rather than a keyword tool, and there is no search volume for a question asked this way. On 28 October 2026 we re-ran the Singapore audit in four languages, 21 batches and 126 questions, producing 42 readings with no failures. The brand batch names us inside its own prompt and is excluded from every figure, which leaves 40 blind readings, 16 of them in English. In those 40 blind readings AI Knows Us was named 0 times and cited 0 times.

"How do I get my products recommended when a buyer asks an AI assistant what to buy?" produced two readings, one on ChatGPT and one on Gemini, and neither named a supplier, a platform partner or a tool. Across the whole blind set the vendors the two engines did name were Otterly in 9 readings and cited in 6, Semrush named in 9 and cited in 10, HubSpot named in 9 and cited in 5, Profound named in 7 and cited in 2, Peec named in 7 and cited in 7, and Ahrefs named in 6 and cited in 5. Named and cited are two separate findings on two separate measures and they are never added together.

The practical reading is that nobody has won this question yet. A Singapore seller asking it today gets a description of the idea, not a shortlist, which means the three catalogue routes above are the whole of the actionable answer and they are all free to read.

The order to do it in, if you sell physical products from Singapore

• Read your own robots file and remove any block on OAI-SearchBot that nobody can account for.
• Open or claim a Merchant Center account and get the free listings feed accepted, because it is the one route that reaches Gemini, Search, Maps, Lens and the Shopping tab from a single submission.
• Add merchant listing structured data to the product template, with price, availability, shipping and returns, matching the feed exactly.
• If you are on Shopify, enable the Catalog and choose your AI channels, which on the Agentic plan costs "$0 SGD /mo" with payment only on a sale, read 29 October 2026.
• Decide whether a ChatGPT catalogue through the Agentic Commerce Protocol is worth engineering time this quarter, and if it is, treat it as a feed project with an owner rather than a marketing project.
• Only then spend money on being described correctly in the written answers, which is a different job from being in the catalogue.

Where a marketplace fits into that order is a separate decision, and the fee side of it is set out in our page on adding Shopee, Lazada, TikTok Shop or Amazon to your own Singapore store, which records what each seller route actually served when we opened it.

Questions about getting a Singapore product named in an assistant's answer

Is this the same work as SEO for my product pages?

It overlaps and it is not the same purchase. Ranking a category page is a content and linking job. Being eligible for a shopping surface is a feed job with a validation report, and the thing that breaks it is a disapproval rather than a competitor. A seller who has both wins twice from one set of product data, which is the argument for treating the feed as the first piece of work rather than the last.

Do I need Shopify for any of this?

No. Merchant Center free listings and product structured data work on any platform, including a WooCommerce or custom build, and so does the robots file check. What Shopify adds is a managed route into its own Catalog for AI channels at no monthly cost, read 29 October 2026, which is a convenience rather than a requirement.

Will a free listing actually put my product into a Gemini answer?

Google's own documentation names Gemini among the places free listings can appear, read 29 October 2026, and lists eligibility requirements and an approval step before anything shows. So the honest way to plan it is as eligibility rather than placement. Get the feed accepted, keep the data clean, and check the performance reporting in the account rather than asking the assistant and reading the answer as a measurement.

How often should a Singapore seller re-check all this?

Monthly for the feed, because disapprovals arrive silently and a product that drops out looks exactly like a product nobody searched for. Quarterly for the documentation, because the operator pages move. The free listings page we read on 29 October 2026 names seven surfaces including Gemini, and a restated operator list goes stale into a false claim unless somebody re-reads it and dates the reading.

What should I ask a supplier who offers to do this for me?

Three things, and they are answerable in one email. Who owns the Merchant Center account, which should be your company with the supplier granted access. Who gets the disapproval email and how often they look. And what the supplier will change on the product template, named field by field, rather than described as optimisation.

Where we sit on this question, and the vendor that beats us on price

We sell the measurement half of this, across five apps, ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. The criterion we ask to be judged on is whether a vendor publishes what it found including the parts that do not flatter it, because that is checkable without a sales call. On that criterion, here is ours: in the 28 October 2026 Singapore re-sweep we were named in 0 of 40 blind readings and cited in 0, and in the 26 October 2026 Singapore run we were named 0 and cited 4. Those are our own numbers about our own position in this market and they are not rounded.

On entry price a rival is cheaper and it is worth saying so plainly. Otterly publishes a Lite tier at "$ 29 /month" for 15 search prompts, read on otterly.ai on 29 October 2026, with four engines tracked and Claude, Gemini and Google AI Mode as paid add-ons. Our own Singapore cards, read on aiknowsus.com on 29 October 2026, start at SGD 194 a month for SEO Start and SGD 389 a month for GEO Growth, shown in Singapore dollars and billed in US dollars. Anybody whose question is purely how little they can spend to start tracking prompts should look at the cheaper tier first. What a seller can take from us at no cost is the reading itself, a dated measurement of what the five assistants say about your products, which is a sensible thing to have in hand before you brief anybody on a catalogue project.