National Trading Standards has published an AI Scam Watchlist that maps how criminals can use synthetic voices, deepfake video, fake websites and automated conversations to make familiar frauds more convincing and scalable.
The September 30 release matters to banks, payment service providers and fraud teams because several of the listed schemes end in a card payment, bank transfer or disclosure of credentials. The document does not identify a new payment-system breach or quantify losses caused by artificial intelligence. Instead, it describes how criminals can manufacture the context that customers and frontline controls often use to decide whether a payment request looks legitimate.
Alongside the watchlist, National Trading Standards reported a survey of 2,000 UK adults conducted by Censuswide from September 18 to 22. Forty-seven percent said it was becoming harder to tell what is genuine online, while 24% said the growth of AI had left them overwhelmed by increasingly sophisticated scams. Another 19% said scam concerns had caused them to avoid some online activity.
Those figures measure attitudes, not verified victimization or payment losses. National Trading Standards also reported that 62% of respondents felt confident they could identify an AI-created scam, but the poll did not test whether they could do so accurately.
AI is scaling old scams rather than creating a new payment rail
The watchlist says AI is accelerating what it calls the industrialisation of fraud. Its evidence base includes National Trading Standards intelligence logs and cases, material from the Chartered Trading Standards Institute and input from partner organizations.
The criminal toolkit it describes includes digital humans, deepfake video, voice cloning, AI-generated websites, messages and images, fake reviews, chatbots and manipulated information that can cause AI search results to point users toward fraudulent sites. National Trading Standards applies those techniques to ten familiar scam categories, including shopping, investment, romance, travel, technical-support and family-impersonation fraud.
The payment path varies by scheme. A fake retailer may collect a card payment and card details. An investment scam may begin with a small transfer, show a fabricated return and then seek a larger payment. A romance or family-impersonation scam may use a synthetic voice or video to create urgency before asking for money. A technical-support scam may seek remote access, passwords or one-time security codes.
This is a control problem because a customer can complete a payment after apparently checking several sources. AI can help criminals surround a false proposition with a polished website, plausible reviews, responsive chat and convincing audio or video. The presence of those signals therefore cannot be treated as independent corroboration.
The watchlist shifts the verification question
National Trading Standards advises consumers to stop before paying, check a request through an independent channel and protect money and information until the recipient is verified. Its category-specific guidance recommends calling a relative back on a trusted number, checking an investment firm through the Financial Conduct Authority, and avoiding bank transfers when paying an unfamiliar retailer or travel seller.
For payment providers, the practical implication is that customer recognition of a voice, face or brand is becoming a weaker assurance signal. Fraud controls should be designed around evidence the criminal does not control: the customer’s established device and behavior, the history and risk of the receiving account, sudden changes in payment amount or destination, and confirmation through a trusted channel.
That does not mean every unusual payment should be blocked. It does mean that warnings and step-up checks should reflect the scam pattern presented by the transaction. A first payment to a new investment beneficiary, an escalating series of transfers, or a payment made during an active remote-access session can carry different risks from an ordinary recurring payment. Providers should also test whether staff and automated interventions remain effective when a customer believes they have seen or spoken to the person requesting the money.
Consumer education is only one layer
The release places substantial emphasis on consumer awareness, but it also records calls for wider accountability. Rocio Concha, Which?’s head of policy and advocacy, said fraudulent advertising should be addressed at its source and called for enforcement against online platforms that fail to comply with applicable obligations. The Advertising Standards Authority encouraged reports of suspected paid scam advertisements so intelligence can be shared with enforcement partners.
For fraud and payments teams, the useful lesson is narrower than a claim that AI has created an entirely new class of crime. National Trading Standards is warning that criminals can now produce more persuasive evidence around a payment request at lower cost and greater scale. Controls built mainly on the assumption that customers will spot poor grammar, an implausible website or an unfamiliar voice are becoming less dependable. The accountable response is to strengthen verification, beneficiary intelligence and intervention at the point money is about to move, while measuring whether those controls stop losses rather than merely generate warnings.