How do scammers use artificial intelligence to target victims?
Scammers use AI tools to generate convincing personalised messages at scale, create deepfake voices and images, and automate the early stages of relationship building.
Last reviewed: 10 June 2026
Explanation
Artificial intelligence has lowered the cost and raised the quality of scam operations. Tasks that previously required skilled human writers or callers can now be automated. Large language models generate grammatically correct, contextually appropriate messages that do not carry the spelling and phrasing errors that were once useful warning signs of overseas-based fraud. Personalisation — inserting a target's name, referencing their job or interests — can be applied automatically to thousands of messages per hour.
Deepfake voice technology allows scammers to impersonate real people in phone calls. Audio from a public social media video of a family member or colleague can be processed to create a synthetic voice capable of saying anything. This enables the grandparent scam to work even when the grandparent asks the caller to confirm something only the real grandchild would know — with enough source audio, even that barrier can potentially be overcome.
Deepfake video, while more computationally demanding, has been used in romance and business contexts. A scammer conducting a video call can use a real-time face-swap tool to display a different identity. In business fraud, executive impersonation via video has been used to authorise large wire transfers in what appeared to be a legitimate video conference with a senior figure from the organisation.
AI also assists with targeting. Natural language processing tools can analyse public social media at scale to identify individuals whose posts indicate emotional vulnerability, financial availability, or relevant life events. The combination of mass data analysis and individually personalised outreach makes AI-assisted fraud significantly more effective than traditional blast approaches.
Common red flags
- Messages are fluent and personalised but feel slightly generic in their warmth
- A phone caller sounds like someone you know but the voice has unusual rhythm or artifacts
- A video call image is slightly off — unnatural movement, blurring at the edges of the face
- The same conversational structure appears in multiple separate contacts over time
- An 'executive' on video authorises an unusual financial instruction but is hard to reach afterwards
- Responses to your messages arrive too quickly and too consistently
What to do now
- Establish a personal verification code with family members for emergency calls
- Ask unexpected callers a question only the real person could answer, including an obscure personal detail
- For large financial authorisations, require a separate verification step outside the same communication channel
- Be sceptical of video calls where lighting, synchronisation, or facial movement seems slightly off
- Report suspected deepfake impersonation to the platform and to law enforcement
- Stay informed about how AI fraud is evolving via consumer protection resources
Frequently asked questions
Can you tell if you are talking to an AI chatbot versus a real person?
Not reliably. Modern language models produce human-like text. Clues like very fast response times, lack of genuine personal detail when probed, and no memory of contradictions can help, but the distinction is increasingly difficult to make without technical assistance.
Are there tools to detect deepfake audio?
Research tools and some commercial services attempt to detect synthetic audio, but the technology is improving faster than detection. The most reliable defence is a verification system agreed in advance with family or colleagues.