Fake reviews in Singapore: what CCS found and how to check

On 30 September 2026 the Competition and Consumer Commission of Singapore (CCS) announced its largest fake-review investigation to date by number of businesses. In its media release, CCS says 45 businesses admitted buying fake online reviews from one provider, which served about 100 businesses and used generative AI to write reviews that read like real customers. The undertakings include removing the fake reviews and displaying a public apology for six months.

For a shopper comparing sellers, the lesson is narrow. A star rating can be manufactured, and a fluent, specific-sounding review does not prove that a real customer wrote it. That does not mean most ratings are fake: the release covers one provider’s clients, not how common fake reviews are. Below is what CCS found about fake reviews in Singapore, and a five-check audit, our own heuristic, for any listing.

How were the fake reviews produced and posted?

CCS names the provider as Reputifly Pte. Ltd. and its websites BuyReviewSG and GetReviewSG, and says it began on or before 2023. According to the release:

• Packages and a calculator. Businesses could buy packages of fake reviews to be posted periodically, and use a rating calculator to work out how many five-star reviews they needed to reach their desired Google rating.
• Generative AI. It was used to vary writing styles and sentence structures, add details such as Singapore context and introduce natural imperfections. A dashboard could draw on a business’s existing four- and five-star Google reviews to generate new ones. Some reviews deliberately described initial reservations before a positive experience.
• Paid posters. The provider recruited people through Telegram, paying them to post prepared content and ratings, and told them to avoid duplicate reviews to reduce the risk of detection.
• Replacements. It offered to replace reviews that platforms removed, capped at 30 per cent of the reviews purchased.
• Platforms. The release says these may include Google, Facebook, Tripadvisor, Carousell, Yelp and Trustpilot.

Our reading, not CCS’s: polish, local references and typos prove nothing, a real-looking reviewer profile proves little because the posters were paid, and a worry that ends in praise is not reassuring on its own.

What must the businesses do, and what is still open?

The 45 businesses undertook to stop, to identify and remove fake reviews relating to them, to report progress to CCS, to strengthen compliance and to publish a prominent apology for six months on their websites, official social media accounts and at their premises or outlets. The provider undertook to stop and to publish an apology prominently on its company’s website for six months, and voluntarily committed to donate its proceeds from the service to an Institution of a Public Character.

Three things were open or unstated when we rechecked on 17 October 2026:

• Two businesses. Two of the 47 declined to give undertakings on CCS’s terms. CCS says it will continue investigating them and take firmer action if the facts support it. The release page does not say whether they also admitted buying reviews, and CCS’s newsroom list showed no further release on the case as of 17 October.
• Phase two. CCS says a second phase extends beyond the remaining businesses to others that may have bought from the provider, and it has asked such businesses and other providers to come forward voluntarily.
• Apology dates. Neither the release nor CCS’s list of the 45 businesses says when each business’s apology period started.

CCS’s Chief Executive, Alvin Koh, said CCS “will pursue not only those who buy them, but also those who profit from supplying them”.

How does this case compare with the two earlier ones?

CCS’s case register describes two earlier cases, written under the older abbreviation CCCS for the same body.

• Loft Home Furnishing, 21 June 2024. Customers complained that reviews with their initials were posted on the company’s website without their knowledge. CCCS found fake five-star reviews that appeared to be made by at least seven customers, published between November 2022 and August 2023. Undertakings included a feedback channel and removing verified fake reviews, and CCCS issued warnings.
• Lambency Detailing, 3 July 2025. Its operator admitted posting fake five-star reviews on its Sgcarmart.com business page for two years, using customers’ information without consent, through a QR code that needed no account, and using ChatGPT to write the text. CCCS called it the first case involving both a third-party platform and AI.
• The provider case, 30 September 2026. An outside supplier sold reviews to about 100 businesses.

Our reading: the cases move from a business posting on its own website, to a business using AI on a third-party platform, to a supplier selling reviews at scale. The Lambency page calls a business posting fake customer reviews in relation to a consumer transaction an unfair trade practice. For platforms and retailers, a joint release of 26 September 2025 says the enhanced Technical Reference 76 for e-commerce includes guidance on addressing fake reviews, which the release defines as fabricated or manipulated reviews that do not reflect consumers’ genuine opinion or experience. The release does not set out what that guidance says.

How can you check a star rating? A five-check review audit

Each check is tied to something CCS found. None is proof alone. Write down what you need first, as in our simple product brief, so the rating is one input and not the whole decision.

1. Timing: look for bursts. Sort by newest and count reviews per week. A cluster of five-star reviews in a short window deserves a closer look. CCS says packages were posted periodically, so an even timeline is not proof that reviews are genuine, and a burst can have innocent causes such as a promotion.
2. Wording: do not reward polish. Ask whether a review says something you can check: the exact item or service, roughly when, what went wrong and how it was fixed. Watch for one story shape repeating.
3. Purchase signal: who may review? CCS’s dark-patterns page gives, as one safeguard a platform may use, ensuring that only those who purchased can leave a review. Check whether the platform shows a review comes from a buyer. Our reading: reviews on a seller’s own website are published by the seller, as in the Loft Home case above, so they say little about independent buyers. Where nothing shows a purchase, treat stars as claims.
4. Rating shape: read the middle and the bottom. The same page describes deleting negative reviews from genuine customers. Read the one- to three-star reviews and the seller’s replies, and recompute the average without the newest burst. A rating skewed to five stars can have innocent explanations, so it is a prompt, not a verdict.
5. Apology notice: has the seller said anything? Check the seller’s homepage, social profiles and storefront for an apology about fake reviews. The 45 businesses must display one for six months, and Lambency Detailing’s 2025 undertakings also required six-month notices, so a notice need not come from this case. You can also look the seller up on CCS’s list of the 45 businesses. It gives registered names, which may differ from a shop’s trading name (our reading), so match the name on an invoice or the seller’s website first; CCS itself notes that Sanook Kitchen is not related to the listed Sanook Group Pte. Ltd. No notice and no match tells you almost nothing: CCS gives no start dates and the second phase is open.

What does the audit say about one fictional listing?

This example is illustrative, with invented figures. A fictional aircon servicing company shows 150 reviews: 132 five-star, 8 four-star, 3 three-star, 2 two-star and 5 one-star. Of the five-star reviews, 96 arrived in the last 21 days.

• Share of the total: 96 ÷ 150 = 64 per cent of all its reviews arrived in 21 days.
• Average with everything: (132×5 + 8×4 + 3×3 + 2×2 + 5×1) ÷ 150 = (660 + 32 + 9 + 4 + 5) ÷ 150 = 710 ÷ 150 = 4.73.
• Average without the 96 recent reviews: (710 − 96×5) ÷ (150 − 96) = (710 − 480) ÷ 54 = 230 ÷ 54 = 4.26.

The burst lifted the listing from about 4.26 to 4.73. That proves nothing on its own: a promotion, a campaign asking customers to review or a platform prompt can also produce one. It is a reason to run the other checks. Suppose the platform also shows nothing about whether reviewers bought the service, and the recent reviews are mostly one line of generic praise. Checks 1 to 4 then each raise a flag, four in all.

Our own decision rule, not an official threshold: with 0 or 1 flag, treat the rating as one weak input beside price and terms; with 2 or 3, do not rely on the stars and judge on older reviews, other platforms and written terms; with 4 or 5, shortlist another seller unless you can confirm quality another way. No number of flags shows that reviews are fake; that is for CCS to establish. This listing has four flags, so under our rule these steps are the way to confirm quality, and if they fail, another seller is the better shortlist:

1. Cross-check another platform: do the reviews there match the older 4.26 stretch?
2. Read the 18 older reviews below five stars and how the company replied.
3. Ask for a written quote with terms and a reference number, so your decision rests on commitments the company can be held to.
4. Save screenshots of the listing and quote, as in our checkout records guide.

What can you do if you suspect fake reviews?

CCS’s page on reporting unfair trade practices says you may contact CASE at 6277 5100 about a business engaging in an unfair trade practice, and may share feedback with CCS on egregious cases by filing a complaint. The page was last updated on 26 September 2025, so confirm the details before you call. If your own name or details appear on a review you did not write, mention it when you report: the Loft Home case began with complaints from customers in that position.

Related guides: online shopping scam checks for seller identity and false urgency, the year-end sales price check, the mattress sleep-trial guide for a separate CCS rating-badge case, and the Lemon Law action plan if goods are faulty. More is under comparing.

Common questions