Online scams have always existed. What is changing today is their level of sophistication. Artificial intelligence makes it possible to generate convincing profiles, natural conversations, fake documents of near-professional quality, and video or audio deepfakes that imitate real people. The classic warning signs - spelling mistakes, language inconsistencies, generic photos - are becoming increasingly unreliable.
Understanding how these scams work is a first step. But understanding them is not enough to defend yourself alone.
How AI is used in modern scams
Generative AI tools now make it possible to create profiles from scratch: a realistic photo generated by an image model, a coherent biography, a history of social media posts, even fake identity documents or bank statements. What once took hours of manual manipulation now takes minutes, and the result is often undetectable to the naked eye.
In romance scams, AI enables simultaneous conversations with dozens of potential victims, adapting tone and content to each person. In investment scams, it generates fake performance reports, fake screenshots of gains, and fake satisfied customer testimonials. Audio and video deepfakes make it possible to imitate the voice or image of celebrities or executives to legitimise an urgent request.
New warning signs - and their limits
Faced with AI-generated scams, certain signals evolve. A conversational fluency that is too perfect, responses that seem slightly off in relation to what you just said, an inability to answer very specific questions about personal details, video or audio whose synchronisation seems slightly artificial.
But these signals are subtle, not always present, and AI models improve rapidly. What was detectable six months ago may no longer be today. Relying solely on these signals for protection means playing an unequal game.
What AI cannot falsify
Despite the sophistication of AI tools, certain information remains difficult to falsify at scale. The actual history of a legal entity in independent official registries. Court decisions available in public databases. Consistent digital traces over several years in sources the scammer does not control. The correspondence - or absence of correspondence - between a declared identity and what appears in independent registries.
It is not the detection of AI-generated text that protects - it is the verification of what actually exists behind the person or entity, independently of what they show you.
The most affected sectors
AI-powered scams particularly affect: online investment (crypto, forex, trading), romantic relationships, fake recruitment where an attractive job offer conceals a request for personal data or money, and fake technical support scams where an AI imitates an advisor from a major company. In each of these cases, the sophistication of the scam makes independent verification all the more necessary.
What we do
At YMV & Co., we carry out verifications for private individuals who have doubts about the authenticity of a person, platform or opportunity encountered online. Our approach looks for what AI cannot invent: the real past, independent traces, inconsistencies between what is declared and what is verifiable.