ChatGPT Shopping Research: What Singapore E-commerce SMEs Should Do This Month

If you sell anything online from Singapore, the way buyers find products is splitting in two. One half still goes to Google. The other half is starting in ChatGPT, where the answer is a curated comparison instead of ten blue links. The shopping research feature is now available to free users on web and mobile, and it pulls from "verified" sources to recommend products on price, reviews and specifications.

Most local SME owners I talk to are reacting in one of two ways: ignoring it because their sales still come from Meta ads and Shopee, or panicking because they think they need to rebuild their stack. Both are wrong. There is a small, specific list of things to do this month, and most of it is housekeeping you should have done anyway.

Key Takeaway: ChatGPT now researches products for shoppers using a fine-tuned GPT-5 mini variant and pulls from review sites and structured product data. To show up, fix your product schema, get into the review sites ChatGPT trusts, and stop assuming Instant Checkout will save you — OpenAI quietly deprioritised it in March 2026.

Written by Derek Chua, digital marketing consultant and founder of Magnified Technologies. I help Singapore SMEs build practical marketing systems that survive the shift to AI search.

What Actually Changed in 2026

Two things are worth your attention.

First, shopping research is generally available. ChatGPT now runs a specialised version of GPT-5 mini for product queries. OpenAI reports it hits 52% accuracy on multi-constraint queries like "wireless earbuds under S$200 with good noise cancelling for small ears", versus 37% on standard search. That is still wrong half the time, but the trajectory is clear.

Second, Instant Checkout has been deprioritised. In March 2026, OpenAI publicly walked back the in-chat purchase flow it launched in late 2025 with Etsy and Shopify, saying it "did not offer the level of flexibility we aspire to provide". The Buy buttons are still around in some places, but the strategic centre of gravity has moved back to research and recommendations. Shoppers click out to your site to buy.

This is good news for most Singapore SMEs. You do not need to integrate anything. You need to be findable and trustworthy when ChatGPT does its research.

How ChatGPT Decides Which Products to Recommend

From the documented behaviour, the model leans on three signals:

1. Structured product data — schema markup on your product pages, your feed in Google Merchant Center, and any data you syndicate to marketplaces.
2. Third-party reviews and comparisons — Wirecutter, Reddit threads, YouTube reviews, niche blogs, and aggregator sites that LLMs have trained on.
3. Brand and product mentions in authoritative content — write-ups in publications, founder interviews, and editorial coverage that links your product name to the problem it solves.

If your only digital footprint is a Shopify storefront and Meta ads, ChatGPT has very little to work with.

What Singapore SMEs Should Do This Month

Here is the practical list, in priority order. You can work through this in a week.

1. Audit your product schema. Every product page should have valid Product schema with name, description, brand, offers.price, offers.priceCurrency (SGD), aggregateRating, and review fields. Test in Google's Rich Results Test. If you are on Shopify, most themes generate this automatically — verify it actually outputs. If you are on a custom build, it is often missing entirely.

2. Get your reviews on the platforms LLMs read. That means Google Business Profile reviews, Trustpilot, and your industry-specific aggregators. A handful of recent, detailed reviews is worth more than a thousand stars on a widget that no crawler can reach. If you sell beauty, food, or fashion, also seed product mentions in relevant Reddit communities — not as spam, but as real participation.

3. Write one comparison page per product line. "Best [product] in Singapore 2026" and "[Your product] vs [competitor]" pages have always been good SEO. They are now critical because LLMs cite comparison content when shoppers ask "which is better". Be honest about trade-offs — readers and models both notice when you are not.

4. Check your robots.txt is not blocking AI crawlers. I see this one a lot. A WordPress security plugin or an old GDPR config silently blocks GPTBot, OAI-SearchBot, PerplexityBot, and ClaudeBot. If you want to be in the answers, you have to let them crawl. (See our guide on AI crawler robots.txt rules for the specifics.)

5. Make sure your prices, currency and stock status are machine-readable. "Call for price" pages will not be recommended. Neither will pages where the only price is in a JPEG. List the price as text, in SGD, with a clear in-stock/out-of-stock indicator.

6. Get one credible third-party mention this quarter. A press release in a niche trade publication, a podcast interview, a guest post on an industry blog. One real mention from a site LLMs trust will do more than ten on your own blog.

What Not to Bother With Yet

Skip the in-chat checkout integrations for now. OpenAI's own retreat is a signal that the merchant-side commerce protocol is not stable. Likewise, do not pay for an "AI shopping optimisation" tool until you have done the schema, reviews and comparison content basics. Almost every paid tool in this space is automating things you should understand before you outsource.

I would also be cautious about producing AI-generated product descriptions at scale. The current generation of LLMs are surprisingly good at detecting their own output, and AI-on-AI content tends to get filtered out or down-weighted in shopping research. Write descriptions a human would actually read.

What This Means for Your Business

If you are a Singapore SME with under S$2 million in online revenue, the realistic upside this year is not "ChatGPT becomes my biggest channel". It is "I do not lose share to competitors who got organised earlier". The shoppers who use AI for research are higher-intent — they have already filtered, compared, and shortlisted before clicking through. Conversion rates on AI-referred traffic are reported to be much higher than generic organic.

Your goal is to be on the shortlist. That is a much smaller, more achievable target than "rank #1 on Google".

Working With Magnified

If you want a second pair of eyes on your e-commerce setup — schema, review presence, robots.txt, and the comparison pages worth building first — I offer a free 30-minute consultation. I will give you a prioritised list of what to fix and what to leave alone. No pitch, no upsell.

FAQ

Does my Shopify store work with ChatGPT shopping research automatically? Mostly yes for being indexed, no for purchases. Shopify generates reasonable product schema by default, so your products are eligible to be referenced in research. Instant Checkout, the in-chat purchase flow, has been deprioritised by OpenAI. Shoppers will click out to your site to buy.

Should I add an llms.txt file to my site? Probably not yet, and not as a priority. Google has publicly said its Search team does not use it, and OpenAI has not endorsed it as a ranking signal either. Spend the same hour on schema and review content instead.

How do I know if ChatGPT is recommending my products? Ask it. Open ChatGPT logged out, and run shopping queries a customer would run — "best [your category] in Singapore", "alternatives to [your competitor]", "where to buy [your product type]". Do this monthly. For a more systematic approach, see our AI brand monitoring guide for Singapore SMEs.

Will paying for SEO still matter? Yes. The same content and authority signals that win in Google search are the foundation for being cited in AI search. There is no separate "AI SEO" — it is regular SEO done well, plus a few specific additions like robust schema and credible third-party content.

Is this just a US thing? Will Singapore shoppers use it? Adoption in Singapore is behind the US but rising. The local users most likely to use ChatGPT for shopping research today are professionals making considered purchases — electronics, home goods, gifts, services. If your target customer is on LinkedIn, they are already trying AI search.

What about Perplexity, Gemini and Claude? The same fundamentals apply. Schema, reviews, comparison content and editorial mentions help across all of them. Tooling differs, but the inputs do not.

Related reading: How AI is Changing E-commerce — IBM and NRF 2026 Findings for Singapore SMEs and How to Measure AI Search Traffic in Singapore.