South Africa recorded the highest rate of suspected digital fraud of any African country analysed in TransUnion’s H1 2026 Fraud Trends Update, with 3% of consumer transactions flagged as potentially fraudulent.
A third of South Africans who lost money to digital fraud in the past year say it originated on a legitimate e-commerce or social platform, not an obviously fake one — fraudsters are embedding themselves in spaces people already trust, rather than relying on crude fake sites.
That shift is playing out on social media in particular. Kaspersky survey data found 46% of South African scam victims first encountered the fraud on social media, more than any other channel. And confidence isn’t protection: 88% of South Africans believe they can spot a scam, yet 41% say they’ve actually fallen victim.
The New Face of Online Scams
Fraudsters are impersonating trusted brands, banks and even friends and family, using urgency and emotional manipulation to extract personal information, verification codes or immediate payments. The most common threats facing South Africans right now are fake Marketplace listings, investment schemes promising unrealistic returns, romance scams via direct message, phishing for passwords or OTPs, and impersonation of public figures or customer support teams.
The numbers back this up. Digital banking fraud in South Africa has surged 86% year-on-year around R1.9 billion in losses, driven largely by SIM-swap fraud and AI-cloned voice calls. Regulators have also flagged a wave of impersonation and investment schemes, including one that collected roughly R492-million from about 2 000 South Africans before being shut down. Generative AI is amplifying all of it, producing convincing fake images, cloned voices and polished ads that no longer carry the old red flags.
How Meta is Helping Keep People Safe
Meta has rolled out several tools this year aimed at catching scams before they cause harm, rather than only cleaning up after the fact. On Facebook, the company is testing warnings on suspicious friend requests — flagging accounts with few mutual connections or a profile location that doesn’t match where the account is actually operating from, so users get concrete context before deciding whether to accept.
On Messenger, advanced scam detection now watches for patterns common to fraud, such as suspicious job offers from new contacts, and prompts users to submit the chat for an AI scam review; if a scam is detected, the system recommends blocking or reporting the account. On WhatsApp, a new device-linking warning alerts users when someone tries to link their account to an unfamiliar device, a common tactic used to hijack accounts via fake QR codes or verification requests.
Meta’s AI systems also analyse text, images and surrounding context to catch impersonation of celebrities, public figures and brands, and to flag deceptive links that mimic legitimate sites, directly relevant to the fake celebrity investment promotions circulating in South Africa. Underpinning all of this is a push to tighten advertiser verification, with Meta aiming for verified advertisers to account for 90% of ad revenue by the end of 2026.
On education, Meta’s My Digital World programme, developed with digital-safety experts across sub-Saharan Africa for learners aged 13 to 18 covers scam awareness as part of a broader digital literacy curriculum, alongside resources for parents and educators.
Protecting Yourself
A few habits meaningfully cut risk: never share passwords, OTPs or banking details via direct message; verify pages and sellers independently before paying, since a polished profile is no longer proof of legitimacy; treat celebrity-backed investment offers and “too good to be true” deals with suspicion; enable Two-Factor Authentication on all major accounts; and report suspicious accounts and messages immediately rather than ignoring them.
With South Africa now recording Africa’s highest fraud rate, digital literacy has become as important as digital access, and the companies whose platforms these scams run through have a growing case to answer, locally and specifically, not just globally.