AI can imitate a voice or face. Verify the request independently.

Voice cloning: the family emergency scam

Illustrative scenario, not a documented case: a caller sounds like a relative, claims to be in trouble and asks for urgent money. A cloned voice could support the deception, but the same request can also come from an ordinary impersonator. Independently contact the relative before paying.

The FBI documents voice cloning and other synthetic media in its generative-AI fraud warning. Quality and required audio vary; a single minimum clip length is not a universal rule. Our family-impersonation guide gives a verification plan.

Deepfake video calls

Video can support impersonation of a colleague or authority. Before acting on a financial instruction, use your organization’s verified approval process and a separate contact route. Our business-fraud guide explains payment controls.

Visual and audio defects can prompt further checking, but their absence is not authentication. Ordinary bandwidth or lighting problems can also look suspicious. There is no reliable universal head-turn or hand-gesture test for deciding whether a financial request is safe.

AI-generated phishing at scale

Phishing can be mass-produced or individually targeted. Generative AI can help an operator write and adapt messages, but older techniques remain in use:

• ▸

  The FBI logged more than 22,000 AI-related fraud complaints in 2025, with losses above $893 million.

• ▸

  AI can help produce polished, personalized messages.

• ▸

  An operator can supply public job and company details to make a message relevant.

• ▸

  Message variations can complicate detection; this does not mean all spam filters are defeated.

• ▸

  Chatbots can support conversations, while human operators may still direct the fraud.

Good grammar does not establish a genuine sender. Use technical protections and independent verification together.

How to protect yourself from AI scams

Use verification procedures alongside account and device protections:

1. 1.

   Agree a private family phrase as an additional check; replace it if exposed and still verify unusual requests independently.

2. 2.

   Never authorize financial transactions based solely on phone or video calls — verify through a separate channel.

3. 3.

   Be skeptical of any urgent request for money, regardless of who appears to be asking.

4. 4.

   Limit personal audio and video on public social media — scammers mine these for cloning material.

5. 5.

   For businesses: implement verbal verification protocols for wire transfers. No amount should be authorized by voice or video alone.

6. 6.

   Keep your voicemail greeting generic — detailed greetings provide more audio for cloning.

Why the check matters more than the tell

AI can make a false identity more convincing, but appearances are only part of the claim. A real person can also lie. The practical objective is to verify who is making the request, what authority they have and whether the proposed action belongs to a genuine process.

Use visual clues as reasons to investigate, not as a pass/fail test. A family safe word adds a private check; independent contact reduces reliance on the caller’s presentation. Neither is a guarantee if the word or account has been compromised.

Resistance to verification is a warning, but willingness to wait is not proof of legitimacy. Confirm the specific request through the real organization or person, then use the appropriate payment or approval controls.

Secrecy, pressure to remain on the line and instructions not to contact anyone else are reasons to stop and seek independent confirmation. A scammer can also be patient. The useful safeguard is the verification process, not the emotional reaction to your request.

Visual clues remain useful, but do not let a convincing voice or face authorize a payment. Verify the person and the financial request separately. Our AI romance-scam guide explains why a real face cannot establish honest intentions.

What people actually ask AI assistants about scams

There is a second AI story on this page, and it runs the other way. People now ask AI assistants whether a message, an app or a person is a scam, and the assistant answers by reading pages like this one. We publish the questions Microsoft's AI sent to the Bing index when it cited our pages as the AI Scam Query Index: 180 distinct questions in the first release, weekly since July 2026, with our share of each answer. Most of them are comparisons, which payment app is safer, which dating app checks identity, whether a bank alert is real. Almost none of them are about AI scams by name. The people being targeted by cloned voices are asking about the payment rail and the platform, not the technology.

That is why the guide above stays on behaviour rather than detection, and why each AI technique has its own breakdown built on a real case:

• ▸

  The AI voice scam: voice impersonation and independent verification.

• ▸

  Deepfake job interviews: checks for job seekers, and the distinction between fake recruiters and fake applicants.

• ▸

  The AI manager job scam: a chatbot as the recruiter, a task platform as the trap.

• ▸

  The AI trusted-site scam: when the assistant's answer is the phishing link.

• ▸

  AI romance scams: generated faces, scripted affection, and the money question that never changes.

If an AI scam has already cost you money, act fast with our recovery guide, then see where to report it in your country.

Reviewed 23 September 2026. We corrected overbroad verification and recovery claims during our launch-content review. See the corrections log for the material changes.