E-commerce

The fake reviews came with flaws built in

Singapore’s watchdog found a seller using AI to write five-star reviews with deliberate flaws, at S$219 for 35. In Malaysia, review “jobs” are scam bait: part-time job scams, some built on review tasks, hit 8,911 cases and RM222 million in losses in 2025. The Consumer Protection Act never mentions reviews.

BOE Nexus Research30 min read

Food stalls with photo menus on Jalan Alor, Kuala Lumpur, at night
In July 2024, strangers in a WhatsApp “Google Review group” were promised RM15.90 to give a Kuala Lumpur restaurant five stars. The restaurant had not asked for the reviews. Photo: McKay Savage, CC BY 2.0.

On 30 September 2026, the Competition and Consumer Commission of Singapore (CCS) described the inside of a fake-review factory. A digital-marketing firm, Reputifly, sold five-star reviews through two websites. One package offered 35 over seven days for S$219, about RM695.1 After each sale, generative AI wrote reviews “designed to resemble genuine customer experiences”. It varied the style, added local detail and introduced “natural imperfections”. Some reviews opened with doubts before turning to praise.1

In the Malaysian cases we found, the review is mostly bait. In July 2024, strangers in a WhatsApp “Google Review group” were promised RM15.90 to give Bijan Bar & Restaurant in Kuala Lumpur five stars. The restaurant had asked for nothing of the kind.2 In September 2026, a 36-year-old woman in Perak’s Muallim district lost more than RM41,000 after a WhatsApp job offer of what police called “product evaluation tasks on Shopee”.3

Reviews steer money. The UK competition regulator estimates that up to £23 billion of UK consumer spending a year, about RM124 billion, is potentially influenced by online reviews.4 AI now writes convincing praise in bulk, and people cannot reliably spot it.5 This article covers how fakes are made, what platforms and regulators do, and where Malaysian law stands. Voice-cloning scams are covered in our article on AI scam calls. Shopee and TikTok Shop sales are in AI in e-commerce, and travel reviews in AI in tourism.

The short version

  • In the Singapore competition watchdog’s largest fake-review case, a seller used generative AI to write five-star reviews, at S$219 (about RM695) for 35. Around 100 businesses were caught, and 45 must apologise publicly for six months.
  • Published estimates of the fake share lie between about 7% and 16%, and each measures something different. A UK government model put it at 11–15% in three product categories.
  • People cannot tell AI-written reviews from human ones, even when paid for accuracy. A detector firm, Pangram, says 93% of the AI-written Amazon reviews it found carried the “Verified Purchase” badge.
  • Google says it blocked or removed over 292 million policy-violating reviews on Maps in 2025. That is its own figure, and it measures enforcement, not how many fakes exist.
  • In Malaysia, part-time job scams, some of which promise pay for posting reviews, reached 8,911 cases and RM222 million in losses in 2025. The Consumer Protection Act 1999 never mentions reviews, and no source says the coming E-Commerce Bill will.
S$219for 35 five-star reviews over seven days, written with generative AI, from the seller in the Singapore watchdog’s largest fake-review case (about RM695)CNA (30 September 2026)
11–15%of reviews estimated fake in three product categories on e-commerce sites widely used in the UKUK Department for Business and Trade (25 April 2023)
292 millionpolicy-violating reviews Google says it blocked or removed on Maps in 2025Google (16 April 2026)
8,911part-time job scam cases in Malaysia in 2025, with RM222 million lost; some of these scams promise pay for posting reviewsThe Star (26 January 2026); Malay Mail (13 August 2024)

Inside a review factory

The Singapore case began with complaints and with the commission’s own scanning for leads, said Kho Qin Yao, its director of fair trading practices. Investigators found Telegram channels and two websites, BuyReviewSG and GetReviewSG. Reputifly, which also offered web design and search-engine work, had sold fake reviews as an “add on” since 2023 or earlier.1 The reviews appeared on Google, Facebook, Tripadvisor, Carousell, Yelp and Trustpilot.1

The product was built like software. A “review rating calculator” told a client how many five-star reviews it needed to reach a target rating. A later dashboard generated, managed and tracked the fakes. It drew on real four- and five-star Google reviews for convincing detail. If Google removed reviews, Reputifly’s terms promised to replace up to 30% of those bought.1

People did the posting. The firm recruited on Telegram, paying people to post the reviews from their own accounts.1 The day before the announcement, CNA reported Carousell listings that sought or offered single fake reviews for S$1 to S$5, about RM3 to RM16.1

How the Singapore review factory workedThe fake-review service described by the Competition and Consumer Commission of Singapore, September 2026
  1. 1BusinessA business buys a packageFor example, 35 five-star reviews over seven days for S$219. A calculator showed how many were needed to reach a target rating.
  2. 2Seller’s dashboardAI writes the reviewsGenerative AI varied styles, added Singapore context and “natural imperfections”, drawing on real four- and five-star reviews.
  3. 3BusinessApprovalThe client may editClients could edit the reviews before they were published.
  4. 4Telegram recruitsPaid posters publishPeople recruited on Telegram posted the reviews from their own accounts, at intervals.
  5. 5SellerRemoved reviews are replacedIf Google removed reviews, the seller promised to replace up to 30% of those bought.

Some reviews “deliberately described initial reservations before recounting a positive experience”, CCS said.

Source: CNA (30 September 2026), reporting the CCS statement.

Show the numbers
StepWhoWhatDetail
1BusinessA business buys a packageFor example, 35 five-star reviews over seven days for S$219. A calculator showed how many were needed to reach a target rating.
2Seller’s dashboardAI writes the reviewsGenerative AI varied styles, added Singapore context and “natural imperfections”, drawing on real four- and five-star reviews.
3BusinessThe client may edit (approval)Clients could edit the reviews before they were published.
4Telegram recruitsPaid posters publishPeople recruited on Telegram posted the reviews from their own accounts, at intervals.
5SellerRemoved reviews are replacedIf Google removed reviews, the seller promised to replace up to 30% of those bought.

CCS can compel documents under Singapore’s fair-trading law. “We approached him, and we managed to get the information that we need from him, and more critically, we managed to get hold of his database,” Kho told CNA.1 The database turned one seller into a list of buyers. Of 47 businesses in the first phase, 45 admitted buying reviews. They range from restaurants to hair salons, and Kho said the wider group includes doctors, lawyers and tuition agencies.1

The penalties are public. The 45 must remove the fakes, report progress and keep an apology on their websites, social media and premises for six months. Reputifly admitted aiding and abetting unfair trade practices and will donate its fake-review proceeds to a registered charity. Two firms declined to give undertakings, and a second phase covers the other buyers.1 Kho said buyers are a minority, and that CCS acted so honest firms would not feel forced to post fakes to keep up. He also told consumers to refuse paid offers to post fake reviews, which he said would be “poisoning the ecosystem”.1

Businesses should be under no illusion that paying for fake reviews is acceptable. Nor is it acceptable to profit from facilitating the creation or purchase of fake reviews; fake reviews are not a legitimate business.
Alvin Koh, chief executive, Competition and Consumer Commission of Singapore, CNA, 30 September 2026

AI had appeared in a Singapore case before. In July 2025 the commission found that a car-detailing firm had posted ChatGPT-written five-star reviews on its Sgcarmart page for two years, using customers’ details without their consent. A customer complained after finding reviews posted under her name. The regulator also spotted an “unusually high number of 5-star reviews published on certain days”.6

Nobody knows the true share

Estimates exist, but each counts something different. The most thorough we found comes from the UK government. Researchers trained a model on known fakes and ran it over 2.1 million product reviews on nine platforms. They estimated that 11% to 15% of reviews were fake in consumer electronics, home and kitchen, and sports and outdoors.7

Other figures count what a platform caught. Tripadvisor says about 8% of the 31.1 million reviews submitted in 2024 were fake.8 Trustpilot removed 7.4% of submissions in 2024, up from 6.1%, mostly automatically.9 A watchdog firm, The Transparency Company, judged nearly 14% of 73 million reviews of home, legal and medical services likely fake.10 An older study found that Yelp filtered roughly 16% of restaurant reviews.11

Fake-review estimates: what each one counts
SourceWhat it countsShare
UK government, 2023Modelled share of 2.1 million reviews in three product categories11–15%
The Transparency Company, 2024Reviews judged likely fake, home, legal and medical services~14%
Luca and Zervas, 2016Yelp restaurant reviews filtered as suspicious~16%
Tripadvisor, 2024Submissions detected as fake; company figure~8%
Trustpilot, 2024Submissions removed; company figure7.4%

Source: UK Department for Business and Trade; AP; Luca and Zervas, Management Science; CNBC; The Guardian.7891011

Big round numbers deserve suspicion. The CMA’s own March 2026 press release lists a claim that “up to 50% of online reviews are fake”, credited only to “Truth Engine”, with no method given.12 No study we opened supports a figure that high across the web.

Fakes change what people buy, if they are well written. In the UK experiment, shoppers were 5.3% less likely to buy a product with crude fake reviews and 3.1% more likely with subtle ones. For products over £80, about RM432, subtle fakes raised the chance of purchase by 9.2%.7 The researchers put the yearly harm from fake review text on products alone at £50 million to £312 million, about RM270 million to RM1.7 billion. They call that conservative.7 The CMA warns that AI can make fakes “sound fluent, polished and highly convincing”, which is the subtle kind.12

How paid reviews are made

Most fakes come from a small pool of people. The UK study found that a reviewer shared between products predicts fakes better than the words of a review.7 Researchers who watched the trade’s recruiting groups saw the machinery directly. Between March and October 2020 they found about 23 fake-review groups on Facebook each day. Each had about 16,000 members and 568 review requests a day.13

The payment is a refund. The “reviewer” buys the product, posts five stars, then gets the price back by PayPal, plus fees and sometimes a commission.13 The purchase is real, which is why a “verified” label proves little. Amazon says brokers recruit through websites, social media and encrypted messaging, paying in money, free products or other incentives.14

The businesses most tempted are the weak ones. On Yelp, restaurants were more likely to commit review fraud when they had few reviews or recent bad ones, and chains were less likely. More competition brought more fake negative reviews.11 At Tripadvisor, 54% of the fakes caught in 2024 were businesses boosting themselves and 39% were member fraud. Paid reviews were 4.8%, but “more pernicious”.8 Staff also enlist family. “The employees get their mom, best friend or cousin to submit reviews, mentioning their names,” said Becky Foley, Tripadvisor’s head of trust and safety.8

Some tricks need no writer. In “brushing”, sellers ship goods nobody ordered and post reviews in the recipient’s name.15 In “catalogue abuse”, reviews of one product are attached to another. A shopper finds “5-star headphones” whose reviews are mostly about a phone charger, the CMA says.16

Fake reviews sell for about RM3 to RM20 eachPrice per fake review in recent offers, ringgit
Seller of one-star reviews (NYT): US$100 for 20RM20.45
Singapore package: 35 five-star for S$219RM19.86
Malaysian scam “task” pay (a lure)RM15.90
Singapore Carousell listings, topRM15.87
Singapore Carousell listings, bottomRM3.17

Converted at Bank Negara Malaysia’s 17 November 2026 middle rate; per-review prices are our arithmetic. The Malaysian figure is a scam lure, not a market price: the money is made from victims’ deposits.

Source: The Star (The New York Times, 12 September 2025); CNA (30 September 2026); Malay Mail (1 August 2024).

Show the numbers
OfferRM per review
Seller of one-star reviews (NYT): US$100 for 20RM20.45
Singapore package: 35 five-star for S$219RM19.86
Malaysian scam “task” pay (a lure)RM15.90
Singapore Carousell listings, topRM15.87
Singapore Carousell listings, bottomRM3.17

What generative AI changed

AI cut the cost of writing. In 2024 the US Federal Trade Commission (FTC) alleged that paying subscribers to Rytr, an AI writing tool, “could generate an unlimited number of detailed consumer reviews”. Some produced “hundreds, and in some cases tens of thousands” of reviews.17 The Transparency Company said AI-written reviews began to “show up in large numbers in mid-2023”. It had a “high degree of confidence” that 2.3 million of the reviews it studied were partly or entirely AI-generated.10 On Yelp, many AI reviews appeared to be posted by users seeking enough reviews for an “Elite” badge, Pangram’s Max Spero told AP.10

Readers cannot catch them. In two experiments, Balázs Kovács found that people could not recognise AI-written reviews, even with money at stake. They mistook AI reviews for human ones more often than the reverse, and AI detectors were fooled too.5

Detector firms have tried to count. Pangram, which sells an AI detector, scanned 30,000 front-page reviews on 500 Amazon best-sellers. It says 3%, or 909 reviews, were AI-generated “with high confidence”.18 It also says 74% of those AI reviews gave five stars, against 59% of human reviews, and that 93% carried Amazon’s “Verified Purchase” badge.18 Originality.ai, another detector firm, says the likely-AI share of long Google reviews (100 words or more) of US law offices reached 34.4% in 2025, up from 2.04% in 2022.19 These are vendor claims, measured with each firm’s own tool.

AI-written reviews lean towards five stars, a detector firm saysShare of front-page Amazon reviews at each rating, human-written and AI-written, %
Human-writtenAI-written
Five stars59% → 74%
One star22% → 10%

Vendor study: Pangram scanned 30,000 front-page reviews on 500 Amazon best-sellers with its own detector and flagged 909 as AI-written. Not peer reviewed.

Source: Pangram (4 May 2026).

Show the numbers
ItemHuman-writtenAI-written
Five stars59%74%
One star22%10%
Most AI-written Amazon reviews carried the “Verified Purchase” badgeTwo findings from one detector firm’s scan of front-page Amazon reviews
93%of the AI-written front-page reviews it found had the “Verified Purchase” badge
3%of all 30,000 reviews scanned were AI-generated “with high confidence” (909 reviews)

Vendor claims, measured with Pangram’s own detector. A verified purchase can still carry a paid review: in fake-review groups, the buyer is refunded after posting five stars.

Source: Pangram (4 May 2026); He et al., PNAS (15 November 2022).

Show the numbers
ShareWhat
93%of the AI-written front-page reviews it found had the “Verified Purchase” badge
3%of all 30,000 reviews scanned were AI-generated “with high confidence” (909 reviews)
Likely AI-written reviews of US law offices rose after ChatGPTShare of Google reviews of 100 words or more of US lawyers’ offices flagged as likely AI-written, %
0.0%10.0%20.0%30.0%40.0%
34.4%
202020212022202320242025

Vendor claim, measured with Originality.ai’s own detector on long reviews only, from 4,706 collected. AI-written is not the same as fake: some clients may use AI to polish genuine reviews.

Source: Originality.ai (1 November 2025).

Show the numbers
YearLikely AI-written
20200.0%
20210.0%
20222.0%
202313.6%
202426.4%
202534.4%

AI-written is not the same as fake. Foley says most reviews written by AI are not fake, calling the idea that they are “one of the myths I love to bust”. Most come from travellers polishing their prose. Tripadvisor still removed more than 200,000 suspected AI reviews in 2024.8 Some non-native English speakers use AI to get their wording right, AP reported. Sherry He, a marketing professor at Michigan State University, says platforms should target the behaviour of bad actors instead of discouraging honest users.10

Platforms have split. Yelp says reviews must reflect firsthand experiences, “not content drafted or revised by third-party AI tools”.20 Amazon and Trustpilot allow AI-assisted reviews that reflect a genuine experience.10 Pangram told AP that AI reviews tend to be longer and highly structured, with “empty descriptors” and clichés such as “the first thing that struck me”.10

How platforms fight back

The big platforms report rising enforcement. Google says it blocked or removed over 170 million policy-violating reviews in 2023, more than 240 million in 2024 and over 292 million in 2025.212223 In 2025 it also removed over 13 million fake Business Profiles and restricted more than 782,000 accounts.23 In April 2025 it said it had begun alerting users in the US, UK and India when it had removed suspicious five-star reviews, with a global expansion to follow.22 Amazon says it blocked more than 250 million suspected fake reviews in 2023 and more than 275 million in 2024.1424 These are the companies’ own figures. They count enforcement, not how many fakes exist.

Google’s removals of bad reviews rose 72% in two yearsPolicy-violating reviews Google says it blocked or removed, millions
170+2023
240+2024
292+2025

Google’s own figures, each given as “over” or “more than”. They measure enforcement, not the number of fakes. The 72% rise is our arithmetic.

Source: Google (13 February 2024; 7 April 2025; 16 April 2026).

Show the numbers
YearMillions of reviews
2023170+
2024240+
2025292+

Behaviour gives fakes away more reliably than prose. Google’s 2023 algorithm watches for signals such as one reviewer leaving the same review on several businesses, or a sudden spike in one- or five-star reviews.21 Using those patterns, Google says, it caught over five million fake review attempts linked to one task-scam network “in just a few weeks”.21 Amazon’s models weigh links to other accounts, sign-in activity and review history.24 The PNAS researchers found the same. Their model scored 0.890 using who reviewed what, against 0.857 for review text, on a measure where 1 is perfect and 0.5 is a coin toss.13

Who reviews what spots fake-review buyers better than what reviews sayAccuracy of a classifier detecting Amazon products that bought fake reviews, by the data it used, AUC (1 = perfect, 0.5 = chance)
All features combined0.932
Reviewer network0.890
Metadata (timing, ratings)0.874
Review text0.857
Images0.592

Out-of-sample results. The network features record which products share reviewers.

Source: He, Hollenbeck, Overgoor, Proserpio and Tosyali, PNAS (15 November 2022), Table 2.

Show the numbers
Data usedAUC
All features combined0.932
Reviewer network0.890
Metadata (timing, ratings)0.874
Review text0.857
Images0.592

Tripadvisor runs three stages. In 2024 it auto-rejected 7% of submissions, flagged 5% for human review and had staff moderate more than 4.2 million reviews.8 Offenders’ rankings are penalised for a year, and repeat offenders get a red badge “that says this property is trying to deceive you”. “We are the first to admit that we’re never going to reach absolute perfection,” Foley said. “We might not catch [a fake] the first time, but we’ll catch it eventually.”8 Yelp says it filtered out nearly 500,000 suspected AI-generated reviews in 2025.20 Booking.com takes reviews only from guests who booked, within 90 days of checkout.25 Rivals also sue together. In 2025 Amazon and Tripadvisor filed a joint civil action in Milan against a site that allegedly offered reimbursements for five-star reviews.24

One-star reviews as a weapon

Fake reviews now cut both ways. In 2025 The New York Times reported scammers flooding small businesses with fake one-star reviews on Google Maps, then demanding payment to remove them.26 One US contractor paid twice, US$150 and then US$100. Her rating fell to 3.6 from 5.0 before Google removed the reviews. Kay Dean, a former federal criminal investigator who runs Fake Review Watch, identified more than 150 businesses worldwide hit with fake negative reviews.26 One person contacted by the newspaper said he charged US$100, about RM409, for 20 reviews, and denied extorting anyone.26

In June 2026 the Singapore electronics chain Gain City received a flurry of one-star Google reviews. Two to three days later, staff got WhatsApp offers to remove them. The messages went to showroom numbers listed on its Google profile.27 In the US case, taking a mobile number off the business pages ended the WhatsApp contact.26 In April 2026 Google said that after a sudden spike in spam reviews it will remove them, pause new reviews, alert the owner and show customers a banner. These protections were “rolling out globally in the coming weeks”.23

AI summaries lean positive

Platforms now summarise reviews with AI. Amazon says its “review highlights”, launched in August 2023, draw only on reviews from verified purchases.28 A study of 905 Tripadvisor hotels in 21 US cities found AI summaries far more upbeat than the reviews beneath them. On a sentiment scale where minus 1 is most negative and plus 1 most positive, summaries averaged 0.94 and the 96,630 reviews 0.71.29 The gap was widest for weaker hotels and those with few reviews, “exactly where consumers need accurate information most”. The work, by Joanna Quan and Wenqi Shen of Virginia Tech, is in conference slides and has not been peer reviewed.29

How the law caught up abroad

The US wrote a dedicated rule. On 14 August 2024 the FTC voted 5-0 for a final rule banning fake reviews. It covers reviews that “misrepresent that they are by someone who does not exist, such as AI-generated fake reviews”.30 The rule took effect in October 2024. It lets the FTC fine businesses and individuals, but platforms hosting reviews are shielded under US law.10 On 22 December 2025 FTC staff warned 10 companies, citing civil penalties of up to US$53,088 per violation, about RM217,000.31 The same day, the commission voted 2-0 to set aside its 2024 order against Rytr, saying it “unduly burdens innovation in the nascent AI industry”.32

Britain put duties on platforms. Since April 2025, fake reviews, undisclosed paid reviews, hidden negative reviews and misleading star ratings are “banned practices” under the Digital Markets, Competition and Consumers (DMCC) Act 2024. The CMA can decide breaches without a court and fine up to 10% of global turnover.12 In January 2025 Google agreed to put warning alerts on UK businesses caught boosting reviews and to block them from new reviews. Repeat fake reviewers are banned wherever they live.4 In June 2025 Amazon agreed that offending sellers can be banned, and that fake reviewers risk losing all their past reviews.16

In March 2026 the CMA opened five investigations. They include whether Just Eat’s ratings system inflated some restaurants’ and grocers’ star ratings, and whether a pasta company traded discounts for undisclosed five-star reviews.12 The CMA says no finding has been made, and in September 2026 the cases were still open.33 A softer route has worked too: of 54 review publishers sent advisory letters, 90% made changes.12

The EU has banned buying, selling or submitting fake reviews to promote products since 28 May 2022. In a sweep of 223 major websites, reported in January 2022, at least 55% potentially broke EU law. At 144 of them, authorities could not confirm that traders did enough to keep reviews authentic.34 On 11 November 2026 the European Commission proposed a Digital Fairness Act that bans fake reviews, with Commission fines of up to 6% of worldwide turnover for EU-wide breaches. It would apply 12 months after entering into force.35 China’s revised Anti-Unfair Competition Law, in force since 15 October 2025, makes helping others fabricate reviews explicitly illegal. It also covers hiring fake reviewers or reputation agencies to “inflate or tarnish” a product’s reputation.36 Singapore treats false reviews as an unfair trade practice under its Consumer Protection (Fair Trading) Act.1

Of six jurisdictions, only Malaysia has neither a fake-review rule nor a caseHow six jurisdictions treat fake reviews, on sources we could open, November 2026
YesPartlyNoUnknown
ProductExplicit ban on fake reviewsAI-written fakes namedDuties on platformsMaximum penaltyPublic action, 2024–26
United StatesFTC ruleYesin force Oct 2024Yesnamed in the ruleNoplatforms shieldedUS$53,088 per violationPartly10 warning letters
United KingdomDMCC Act 2024Yesbanned practice, Apr 2025Nocovers all fakesYesGoogle, Amazon undertakings10% of global turnoverYes5 investigations, no finding yet
European UnionDirective; DFA proposedYessince May 2022UnknownYesmust say how reviews are handled6% of turnover (proposed)Partly2022 sweep of 223 sites
ChinaAnti-Unfair Competition LawYesArt. 9, Oct 2025UnknownUnknownUnknownUnknownlaw new
SingaporeFair trading lawPartlygeneral unfair-practice lawPartlyapplied to AI cases, 2025 and 2026UnknownCourt orders, restitutionYesabout 100 firms, 2026
MalaysiaAct 599; 2024 e-trade rulesNogeneral misleading-conduct law onlyNoPartly2024 rules, none on reviewsRM250,000 company; RM100,000 or 3 years personNono case found

Our reading of laws and official statements. “Unknown” means our sources did not say, not that no rule exists. Malaysia’s penalties are those of Act 599, section 25, which has not been tested on reviews.

Source: FTC (14 August 2024; 22 December 2025); AP (24 December 2024); CMA (24 January 2025; 27 March 2026); European Commission (20 January 2022); Ashurst Perkins Coie (13 November 2026); China Briefing (3 July 2025); CNA (3 July 2025; 30 September 2026); Act 599; P.U. (A) 449/2024.

Show the numbers
ProductExplicit ban on fake reviewsAI-written fakes namedDuties on platformsMaximum penaltyPublic action, 2024–26
United StatesYes (in force Oct 2024)Yes (named in the rule)No (platforms shielded)US$53,088 per violationPartly (10 warning letters)
United KingdomYes (banned practice, Apr 2025)No (covers all fakes)Yes (Google, Amazon undertakings)10% of global turnoverYes (5 investigations, no finding yet)
European UnionYes (since May 2022)UnknownYes (must say how reviews are handled)6% of turnover (proposed)Partly (2022 sweep of 223 sites)
ChinaYes (Art. 9, Oct 2025)UnknownUnknownUnknownUnknown (law new)
SingaporePartly (general unfair-practice law)Partly (applied to AI cases, 2025 and 2026)UnknownCourt orders, restitutionYes (about 100 firms, 2026)
MalaysiaNo (general misleading-conduct law only)NoPartly (2024 rules, none on reviews)RM250,000 company; RM100,000 or 3 years personNo (no case found)

Malaysia: reviews as currency and as bait

Malaysians buy online at scale. Business-to-consumer e-commerce income reached RM336.6 billion in 2023, the Department of Statistics (DOSM) reported.37 Yet we found no official estimate of how many reviews on Shopee, Lazada, TikTok Shop, Google Maps, GrabFood or Agoda are fake. Nor did we find Malaysian data on rigged food-delivery or hotel ratings. What exists is platform policy, police figures, consumer complaints and one religious ruling.

What the platforms say

Shopee says only buyers of a product can review it. “Shopee takes a serious view of fake ratings or reviews,” a spokesperson told The Star in June 2025. Users found manipulating ratings face frozen accounts or permanent bans.38 Lazada says it uses AI to prompt reviewers to cover durability, design and functionality. It gives its highest LazCoins rewards only to reviews with minimum word counts and photos or video.38 TikTok Shop says it rejected 20.4 million attempted listings and two million seller registrations worldwide in the first half of 2024.39

Users see the gaps. A Kuala Lumpur auditor told The Star of reviews saying “Looks pretty, haven’t tried it yet”, “often repeated in different languages”. An online video-game seller complained that a brand-new rival listing showed over 500 reviews. To help new sellers, Shopee says, it lends new listings reviews of similar items from other sellers, marked “Similar Product”, for seven calendar days.38

This is a big issue. It’s unethical, and more importantly, it misleads others. Some platforms give coins or points for every review, which unintentionally encourages this behaviour.
Azrul Zafri Azmi, deputy president, Malaysia Cyber Consumer Association, The Star, 23 June 2025

Azrul Zafri also says that tying rewards specifically to five-star ratings “becomes manipulative”.38 Most reported scams happen off the platforms. Datuk Armizan Mohd Ali is the Domestic Trade and Cost of Living Minister. He says a large majority of reported e-commerce scams take place after “bad actors redirect consumers off-platform to complete transactions”.39

A religious ruling on review coins

On 18 April 2025 the Federal Territory Mufti’s Office answered a question about Shopee buyers. Some rate and review products as soon as they arrive, before using them, to earn coins. In its guidance, Irsyad Hukum Siri ke-895, it said that rating and reviewing a product without using it deceives and misleads others, and is haram, or forbidden. Shopee Coins received from false ratings and reviews are haram and should be left to expire, it said.40 The office said Shopee’s award of coins is itself permissible under a ju’alah contract, in which one party promises a reward for a specified act. It advised all online buyers to be honest in every review.40

When the review is the bait

Review tasks are a recurring hook for job scams here. In August 2024, Bukit Aman’s Commercial Crime Investigation Department said a scam offered part-time work writing Google reviews, with pay “ranging from RM15.90 per day to RM6,188 per month”.41 The Bijan invitation fits that pattern. The restaurant warned the public that “we never ask for payment or offer incentives for reviews”.2 In a case reported in March 2023, a preschool teacher’s assistant in Pahang lost RM30,483 after answering an offer to review restaurants for commission. She paid deposits into four bank accounts.2 In June 2024 police received 19 reports about a separate syndicate that advertised fake hotel-booking jobs through the LynxOasis.com website, with losses of RM688,099.41

A hotel sign showing a four-star rating
Bukit Aman said in August 2024 that fake Google-review “jobs” promised pay ranging from RM15.90 a day to RM6,188 a month. Photo: Eric Fischer, CC BY 2.0.
How a review “job” scam works in MalaysiaThe pattern described by police and in reported cases, 2023–2026
  1. 1SyndicateAn invitation arrivesA WhatsApp or Telegram message offers paid tasks, such as rating a business on Google or “evaluating” Shopee products.
  2. 2RecruitA small first taskThe first task promises a small sum, such as RM15.90 for giving a restaurant five stars.
  3. 3SyndicateFees and depositsVictims are asked for upfront payments, described as training fees, deposits or permits.
  4. 4VictimDemands growIn one 2026 case, a victim made six transfers totalling over RM41,000, then was asked for over RM32,000 more.
  5. 5NSRCApprovalCall 997 fastThe Home Ministry urges victims to call the National Scam Response Centre on 997 within 24 hours, so funds can be traced and blocked.

Police do not publish how many job scams involve review tasks.

Source: Malay Mail (1 August 2024; 13 August 2024); The Star (26 January 2026; 17 November 2026); The Vibes (17 September 2026).

Show the numbers
StepWhoWhatDetail
1SyndicateAn invitation arrivesA WhatsApp or Telegram message offers paid tasks, such as rating a business on Google or “evaluating” Shopee products.
2RecruitA small first taskThe first task promises a small sum, such as RM15.90 for giving a restaurant five stars.
3SyndicateFees and depositsVictims are asked for upfront payments, described as training fees, deposits or permits.
4VictimDemands growIn one 2026 case, a victim made six transfers totalling over RM41,000, then was asked for over RM32,000 more.
5NSRCCall 997 fast (approval)The Home Ministry urges victims to call the National Scam Response Centre on 997 within 24 hours, so funds can be traced and blocked.

The numbers are large, though no source says how many involve reviews. Part-time job scams reached 8,911 cases in 2025, up from 3,656, and losses rose to RM222 million, up from RM126 million, Bukit Aman said.42 Its director, Comm Datuk Rusdi Mohd Isa, said victims are lured with “simple tasks such as ‘liking’ or ‘following’ content”. A common tactic is to demand upfront payments.42 The average loss per case fell to about RM24,900 in 2025, against about RM34,500 in 2024, by our arithmetic.

The Muallim case shows the pattern. The woman got the offer on 13 September 2026, made six transfers worth more than RM41,000 and stopped when asked for over RM32,000 more. Police are investigating under Section 420 of the Penal Code, which covers cheating. The district police chief urged the public to check job and investment offers before transferring any money.3

Across all types, online scams cost Malaysians RM6.95 billion between 2023 and August 2026, the Home Ministry said in a written reply dated 16 November. E-commerce scams were the second most common type, with 47,221 cases.43 Between January and August 2026, the National Scam Response Centre’s 997 line logged RM250.35 million in reported losses. RM6.67 million was returned to victims, under 3% of those losses, by our arithmetic.43

Malaysia’s online scam losses more than doubled in two yearsLosses to online scams of all types reported to police, RM billion
RM1.22bn2023
RM1.57bn2024
RM2.97bn2025
RM1.18bnJan–Aug 2026

All online scam types, of which job and e-commerce scams are two. The 2026 figure covers eight months only.

Source: The Star (17 November 2026), citing a Home Ministry written reply of 16 November 2026.

Show the numbers
PeriodRM billion
2023RM1.22bn
2024RM1.57bn
2025RM2.97bn
Jan–Aug 2026RM1.18bn

What Malaysian law says, and does not

No Malaysian law names fake reviews. The Consumer Protection Act 1999, Act 599, bans conduct that misleads the public about goods or services, in general terms (section 9).44 Section 10 bans false or misleading representations. These include that “a particular person has agreed to acquire the goods or services”, or that they have “any sponsorship, approval, endorsement, performance characteristics, accessories, uses or benefits”.44 Section 10(1)(h) is the nearest hook for a bought review. The Act never uses the word “review”. In the reprint current to 1 November 2021, the latest amendment took effect on 1 October 2019.44

As far as our sources show, whether that hook reaches a bought or AI-written review is untested. We found no KPDN prosecution or compound for a fake review, and no Malaysian court case on one. Section 25 sets the penalties for breaching these sections. A company faces a fine of up to RM250,000, or RM500,000 for a repeat offence. For a first offence, an individual faces up to RM100,000, up to three years in prison, or both.44

The 2024 online-trading rules do not fill the gap. The Consumer Protection (Electronic Trade Transaction) Regulations 2024 came into force on 25 December 2024.45 Marketplace sellers must disclose set information, such as their name, price and delivery time. They must not provide information they know or have reason to believe is false or misleading. Operators must check that sellers comply, offer a complaints channel and keep seller records for three years.45 The regulations do not mention reviews or ratings.

A new law is coming, but its reach on reviews is unknown. In August 2026 Armizan criticised the Electronic Commerce Act 2006. “This Electronic Commerce Act is merely a legal framework to validate online and electronic commerce, but the aspect of accountability is not clearly outlined in the legislation,” he said.46 In July he had spoken of “at least 10 key policy proposals” for Cabinet, drawn from recommendations by the Malaysia Competition Commission (MyCC) and independent consultants.47 KPDN recorded 8,475 e-commerce complaints in 2025, down from 11,463 in 2021, most commonly about late deliveries and goods unlike their descriptions.47 On 9 October 2026, tabling Budget 2027, Prime Minister Anwar Ibrahim said an e-commerce bill would go to the next parliamentary sitting.48 It rests on four areas: consumer protection; the duties of marketplace operators, vendors and consumers; dispute resolution; and alignment with international model laws. A licensing regime for platform operators, including overseas ones, is also proposed.49 No source we found says it will cover fake reviews.

Malaysian rules and fake reviews, November 2026
RuleWhat it doesReviews
Act 599, section 9Bans conduct that misleads the public about goods or servicesNot mentioned
Act 599, section 10(1)(e) and (h)Bans false claims that a person agreed to buy, or of approval or endorsementNot mentioned; untested on reviews
Act 599, section 25Up to RM250,000 for a company (RM500,000 if repeated); RM100,000, three years in prison or both for a personApplies to breaches of sections 9 and 10, among others
E-trade regulations 2024Seller disclosures, complaints channel, three-year seller recordsNot mentioned
E-Commerce Bill (pending)Platform licensing, duties of operators, vendors and consumers, dispute resolutionUnknown

Source: Act 599 (reprint as at 1 November 2021); P.U. (A) 449/2024; Bernama (14 October 2026).444549

The arguments that remain

Rising removal counts do not prove rising fakes. Foley said Tripadvisor’s doubled detections since 2022 did not mean fakes had doubled, partly because its policy on incentivised reviews had tightened.8 Google’s series is company data with no independent audit that we could find. Amazon has argued that outsiders lack “access to data signals that indicate patterns of abuse”.10 That also means outsiders cannot check its figures.

Policing fakes must not silence honest critics. In Singapore, a household appliance retailer sued a customer for defamation over a negative Shopee review. It lost and must pay S$4,000 in costs, about RM12,700.50

Ratings and reviews are necessary for the efficacious operation of such e-commerce platforms. Sellers who use e-commerce platforms must accept ratings and reviews on their goods and services as part of conducting business online.
District Judge Chiah Kok Khun, Singapore State Courts, judgment published 31 August 2026, reported by CNA, 1 September 2026

The line is fact against opinion. In another case, a TCM physician won S$12,500, about RM39,700, over a patient’s second Google review, which stated serious allegations as fact. The patient’s first review, an honestly held opinion, was protected as fair comment.51 We found no equivalent Malaysian judgment, so these cases describe Singapore law only.

Things you probably didn’t know

  1. Three- and four-star reviews are rarely fake. The CMA advises: “It’s unlikely that a 3 or 4-star review is fake.” Someone might knock off a star because delivery was a day late, it notes.12
  2. “We moderate reviews” banners did nothing. In the UK experiment, telling shoppers that a platform moderated reviews made no significant difference to what they chose.7
  3. Paid reviews come mostly from Asia. Asia supplied only 17% of Tripadvisor’s genuine reviews in 2024. Yet more than a third of the paid fakes it caught came from Indonesia and Vietnam.8
  4. Fake-review buyers cluster. In one Amazon category, two clusters holding just 3.4% of products contained about 70% of the products identified as buying fake reviews.13
  5. A café that becomes a plumber gets flagged. Google trained a model with the help of Gemini to catch suspicious profile edits. Renaming “Zoe’s Coffee House” to “Zoe’s Cafe” is fine; switching category from “cafe” to “plumber” is not.22
  6. Scam firms buy stars too. Suspected scam investment firms used five-star Trustpilot reviews to look legitimate. One rated 4.7 fell to 2.8 after The Guardian flagged it.9
  7. Outrage brings review floods. Yelp says it removed over 63,100 reviews in the US after viral posts in 2025, and alerts tied to politics rose 96%.20

What worked and what didn’t

The fixes that worked went after the sellers and the patterns. The ones that failed asked shoppers or labels to do the work.

Worked

Seize the seller’s records

Singapore’s CCS, which can compel documents, obtained the fake-review provider’s database. One seller became around 100 identified buyers. Of the first 47, 45 admitted buying and must apologise publicly for six months; the seller gives its proceeds to charity.1

Why it worked: a few sellers serve many buyers, so one database exposes them all.

Worked

Deals backed by big fines

With fines of up to 10% of turnover coming from April 2025, the CMA secured undertakings from Google in January 2025 and Amazon in June 2025.416 Of 54 review publishers sent advisory letters, 90% made changes.12

Why it worked: a credible penalty made negotiated fixes quick.

Worked

Watching behaviour, not words

Google says pattern signals, such as identical reviews across businesses and sudden rating spikes, caught over five million fake review attempts from one task-scam network in a few weeks.21 In the PNAS study, reviewer networks beat review text at spotting fake-review buyers.13

Why it worked: patterns across accounts are harder to fake than fluent prose.

Unmeasured

Only paying guests may review

Booking.com accepts reviews only from guests who booked, within 90 days of checkout. Since January 2025 scores are weighted towards recent reviews, and reviews are archived after 36 months.25

Lesson: tying each review to a booking raises the cost of faking, but no source we found measures the effect, and it cannot stop real guests being rewarded for praise.

Worked

Suing the brokers

Amazon says it took legal action against 115 fake-review brokers in two years, and that over 40 brokers and sites stopped after its 2024 actions. In 2024 a Milan court shut Realreviews.it and permanently banned its operator from similar activity.24

Why it worked: brokers have websites, domains and payment accounts that courts can reach. The tallies are Amazon’s own.

Worked

Reporting the recruiters

Yelp says alerts tied to review-exchange rings fell 62% in 2025, partly thanks to better software. Account closures linked to rings and suspicious activity rose 49%. It made over 1,020 reports to platforms such as Instagram, Facebook, X, Reddit and TikTok about people trading reviews; 60% led to action.20

Why it worked: fakes are traded in groups on other platforms, so reports reach the source. The figures are Yelp’s own.

Didn’t work

Trusting the verified badge

Pangram says 93% of the AI-written front-page Amazon reviews it found carried the “Verified Purchase” badge.18 Fake-review groups explain why: buyers purchase the product, post five stars and are refunded.13

Lesson: a badge proves a purchase, not an honest review.

Didn’t work

A fake-review checker for shoppers

Mozilla shut Fakespot, the AI review checker it bought in 2023, on 1 July 2025. It said the idea “didn’t fit a model we could sustain”.52 Kovács found AI detectors fooled by AI-written reviews.5

Lesson: outsiders lack the platforms’ account data, and short reviews are hard to classify.

Reversed

Banning the writing tool

In 2024 the FTC acted against Rytr’s AI review generator. On 22 December 2025 it reopened and set aside the final order, 2-0, saying it “unduly burdens innovation in the nascent AI industry”.32 The ban on using fake reviews stands.30

Lesson: rules aimed at the people who pay for and post fakes have proved more durable than rules aimed at the tool.

Didn’t work

Coins for every review

The Malaysia Cyber Consumer Association says coins or points “for every review” encourage empty reviews. The Federal Territory Mufti’s guidance responded to buyers reviewing before use to earn Shopee Coins.3840 Lazada gives its highest LazCoins rewards only to longer reviews with photos or video.38

Lesson: rewarding the act of reviewing invites reviews with nothing in them.

Too early

Help against review extortion

In 2025 a US contractor paid two sets of extortionists before Google removed the fake one-star reviews. “These businesses are being extorted, and Google isn’t doing enough about it,” Kay Dean said.26 In April 2026 Google announced review pauses and warning banners during spikes, rolling out globally.23

Lesson: whether the pauses reach Malaysian businesses is not yet known.

What to do about it

None of this is legal advice. For a specific dispute, ask a lawyer.

For shoppers

  1. Read the words, not the starsStart with three- and four-star reviews, which the CMA says are unlikely to be fake.12
  2. Look for patternsBe wary of similar wording, a burst of five-star ratings in a short period and reviews with little detail. Singapore’s consumer association names the same signs.112
  3. Distrust slick proseIf a review feels “a bit too slick”, look for detail elsewhere before deciding, the CMA advises.12
  4. Don’t rely on badges“Verified Purchase” shows that someone bought the item, not that the review is honest.1318
  5. Cross-checkCompare the same product on Shopee, Lazada and TikTok Shop, watch short video reviews and read forums, as the consumer association suggests.38
  6. Stay on the platformNever move a deal to WhatsApp or a private link; most reported e-commerce scams happen off-platform.39

If you are offered a review “job”

  • Treat any WhatsApp or Telegram offer of pay for ratings or reviews as a scam pattern, as police have described.4142
  • Never pay a fee or deposit to release earnings. If money has gone, call 997 within 24 hours.43
  • If a parcel you never ordered arrives, change your shopping passwords and tell the platform.15

For sellers, restaurants and hotels

  • Don’t buy reviews or tie rewards to five-star ratings. Singapore made buyers apologise publicly, and its consumer association has raised concerns about discounts or gifts offered only for good reviews. The UK is investigating discounts for reviews.112
  • Take personal mobile numbers off your Google Business Profile, report review spikes to Google and make a police report if extortion follows.2627
  • If strangers are paid to praise you, warn customers publicly, as Bijan did.2

For policymakers

  • Decide whether the E-Commerce Bill should name fake, paid and AI-written reviews, as the US, UK, EU and China have done in different ways.12303436
  • Ask platforms for Malaysian enforcement figures. We found none published.
  • Publish how many job scams use review tasks, so prevention can be targeted.
  • Protect honest negative reviews, as Singapore’s courts have.50

What to watch to 2030

  • The E-Commerce Bill, at the next Dewan Rakyat sitting. The prime minister said it would be tabled then. Watch whether it names reviews or ratings; its four announced areas do not mention them.4849
  • Britain’s five investigations, winter 2026 to 2027. The CMA expects its next case updates then; no finding has been made.33
  • Singapore’s second phase. CCS is pursuing the other businesses that bought from the same seller; no date is given.1
  • Google’s UK reporting, to about January 2028. Google reports to the CMA over three years from January 2025, by our arithmetic.4
  • The EU Digital Fairness Act. Proposed on 11 November 2026, it would apply 12 months after entering into force, and the text will likely change.35
  • FTC enforcement after December 2025. Ten warning letters went out on 22 December 2025. Watch whether any leads to a civil-penalty case under the rule.31
  • China’s Article 9 enforcement. The revised law has been in force since 15 October 2025.36
  • Google’s anti-extortion pauses in Malaysia. Announced in April 2026 for global roll-out; watch whether Malaysian businesses see them.23

Sources

  1. CNA, “Around 100 businesses caught buying fake reviews in largest probe by Singapore’s competition watchdog”, 30 September 2026: channelnewsasia.com
  2. Malay Mail, invitations to paid Google food reviews are a scam, 1 August 2024: malaymail.com
  3. The Vibes, woman loses more than RM41,000 in fake Shopee review job scam, 17 September 2026: thevibes.com
  4. UK Competition and Markets Authority, “CMA secures important changes from Google to tackle fake reviews”, 24 January 2025: gov.uk
  5. Balázs Kovács, “The Turing test of online reviews”, Marketing Letters 35, 651–666, 12 April 2024: link.springer.com
  6. CNA, car-detailing firm posted ChatGPT-written fake reviews on Sgcarmart, 3 July 2025: channelnewsasia.com
  7. UK Department for Business and Trade, “Fake online reviews research: executive summary”, 25 April 2023: gov.uk
  8. CNBC, “Here’s how many fake reviews Tripadvisor found on its website in 2024”, 25 May 2025: cnbc.com
  9. The Guardian, scam investment firms exploit Trustpilot reviews, 19 October 2025: theguardian.com
  10. The Star (AP), “The Internet is rife with fake reviews. Will AI make it worse?”, 24 December 2024: thestar.com.my
  11. Luca and Zervas, “Fake It Till You Make It: Reputation, Competition, and Yelp Review Fraud”, Management Science, 11 February 2016: doi.org
  12. UK Competition and Markets Authority, “Fake and misleading reviews: 5 businesses under CMA investigation”, 27 March 2026: gov.uk
  13. He, Hollenbeck, Overgoor, Proserpio and Tosyali, “Detecting fake-review buyers using network structure: Direct evidence from Amazon”, PNAS, 15 November 2022: pmc.ncbi.nlm.nih.gov
  14. Amazon, latest actions against fake review brokers, 31 July 2025 (updated 8 October 2025): aboutamazon.com
  15. The Star (Tribune News Service), “Got a package you didn’t order?”, 11 January 2025: thestar.com.my
  16. UK Competition and Markets Authority, “Amazon gives undertakings to CMA to curb fake reviews”, 6 June 2025: gov.uk
  17. US Federal Trade Commission, crackdown on deceptive AI claims and schemes, 25 September 2024: ftc.gov
  18. Pangram, “Three percent of front-page Amazon reviews are now AI-generated”, 4 May 2026: pangram.com
  19. Originality.ai, AI-generated reviews of US lawyers’ offices, 2022 to 2025, 1 November 2025: originality.ai
  20. Yelp, Trust & Safety Report 2025: trust.yelp.com
  21. Google, “How machine learning keeps contributed content helpful”, 13 February 2024: blog.google
  22. Google, “New ways AI helps keep business information trustworthy”, 7 April 2025: blog.google
  23. Google, “New ways we’re protecting businesses on Maps”, 16 April 2026: blog.google
  24. Amazon Trustworthy Shopping, working with the Coalition for Trusted Reviews, 7 November 2025: aboutamazon.com
  25. Booking.com Partner Help, everything you need to know about Guest Review Scores: partner.booking.com
  26. The Star (The New York Times), “Small businesses face a new threat: Pay up or be flooded with bad reviews”, 12 September 2025: thestar.com.my
  27. The Straits Times, Gain City flooded with one-star Google reviews, 19 June 2026: straitstimes.com
  28. Amazon, customer reviews and generative AI, 14 August 2023: aboutamazon.com
  29. Quan and Shen (Virginia Tech), “Soft Deception by Design: Algorithmic Positivity Bias in AI-Generated Consumer Review Summaries”, slides, 19 March 2026: ftc.gov
  30. US Federal Trade Commission, final rule banning fake reviews and testimonials, 14 August 2024: ftc.gov
  31. US Federal Trade Commission, warning letters to 10 companies under the consumer review rule, 22 December 2025: ftc.gov
  32. US Federal Trade Commission, Rytr final order set aside, 22 December 2025: ftc.gov
  33. UK Competition and Markets Authority, Just Eat consumer protection enforcement case, updated 28 September 2026: gov.uk
  34. European Commission, “Protecting consumers from misleading reviews: 55% of screened websites violate EU law”, 20 January 2022: ec.europa.eu
  35. Ashurst Perkins Coie, “Digital Fairness Act: Brussels takes on consumer enforcement”, 13 November 2026: ashurstperkinscoie.com
  36. China Briefing, China’s Anti-Unfair Competition Law amendments, 3 July 2025: china-briefing.com
  37. The Star, Malaysia’s e-commerce revenue (DOSM), 26 November 2025: thestar.com.my
  38. The Star, “Ratings, reviews and riddles: What’s the deal with online reviews?”, 23 June 2025: thestar.com.my
  39. The Star, TikTok Shop and KPDN #ShopSafe tips, 21 April 2025: thestar.com.my
  40. Jabatan Mufti Wilayah Persekutuan, Irsyad Hukum Siri ke-895, rating and reviewing unused goods for Shopee Coins, 18 April 2025: muftiwp.gov.my
  41. Malay Mail, federal police expose job scam offering pay for fake Google reviews, 13 August 2024: malaymail.com
  42. The Star, steep rise in part-time job scams, 26 January 2026: thestar.com.my
  43. The Star, online scams cost Malaysians RM6.95 billion since 2023, 17 November 2026: thestar.com.my
  44. Laws of Malaysia, Act 599, Consumer Protection Act 1999, reprint as at 1 November 2021 (hosted by KPDN): kpdn.gov.my
  45. Federal Government Gazette, Consumer Protection (Electronic Trade Transaction) Regulations 2024, P.U. (A) 449, 24 December 2024: repositori.kpdn.gov.my
  46. The Edge Malaysia, Armizan on the new e-commerce law, 22 August 2026: theedgemalaysia.com
  47. The Malaysian Reserve, KPDN to present new e-commerce law to Cabinet, 2 July 2026: themalaysianreserve.com
  48. Free Malaysia Today, e-commerce bill to be tabled in next Dewan Rakyat sitting, 9 October 2026: freemalaysiatoday.com
  49. Bernama, government proposes licensing regime for e-commerce platforms, 14 October 2026: bernama.com
  50. CNA, retailer loses defamation suit against customer over negative Shopee review, 1 September 2026: channelnewsasia.com
  51. CNA, TCM physician wins damages over Google review, 5 August 2026: channelnewsasia.com
  52. Mozilla, “Investing in what moves the internet forward”, 22 May 2025: blog.mozilla.org

Platform figures (Google, Amazon, Yelp, Tripadvisor, TikTok Shop) and detector-firm figures (Pangram, Originality.ai) are the companies’ own claims. They measure enforcement or detection, not the true share of fake reviews. Police and Home Ministry scam figures cover all job or online scams, not review tasks alone. Nothing here is legal advice. Conversions use Bank Negara Malaysia’s middle rate of 17 November 2026: RM4.089 per US dollar, RM5.3967 per pound and RM3.1741 per Singapore dollar.

Talk to an engineer

Bring us the question this raised.

BOE Nexus builds AI systems for Malaysian companies. If this article made you wonder how it would work in your own operation, send us the details. The first consultation is free.

kai@boenexus.com +60 18-403 3314 Petaling Jaya, Selangor, Malaysia