Ukrainian outlet Speka has documented three distinct scam patterns now targeting teenagers online: fake video-game testing job offers, AI-assisted grooming conversations, and deepfake imagery used for blackmail and extortion.
None of these tactics are new in isolation — job-offer scams and online grooming predate generative AI by decades. What's changed is the tooling. Chatbots that can hold a convincing, personalized conversation at scale, image generators that can produce a fake screenshot or explicit image in seconds, and voice cloning that needs only a few seconds of audio have turned what used to require real human effort into something a single bad actor can automate against hundreds of targets at once.
According to Speka, these three vectors now form a recognizable pattern that parents, educators, and platform operators should treat as a single connected threat rather than three unrelated scam types.
Three vectors, one target demographic
Each tactic exploits a different weak point in how teenagers use the internet, but they share a common playbook: build trust quickly, isolate the target from adult oversight, and extract something of value — money, images, or personal data.
- Fake game-testing offers — scammers pose as recruiters or studios offering paid access to test unreleased games, luring teenage gamers with the promise of early access and pocket money, then pushing them toward malicious downloads or requests for payment/personal details.
- AI-assisted grooming — predators use AI-generated personas and chatbot-style conversation to build rapport with minors faster and more convincingly than a human typing alone could manage, then escalate toward isolation and exploitation.
- Deepfakes — fabricated images or video, often sexualized or otherwise compromising, are used to intimidate, humiliate, or extort teenagers, including threats to share the fake content with family or classmates unless demands are met.
Why generative AI lowers the barrier
The practical shift for anyone building or moderating consumer platforms is economic, not conceptual. Grooming and sextortion schemes used to require sustained, skilled human effort per victim — writing convincing messages, building a believable backstory, producing fake images by hand. Generative tools compress that effort. A single operator can now run dozens of AI-scripted conversations in parallel, and image generation removes the need for any real photo of the victim to produce a convincing threat.
That scalability is the part most relevant to AI builders: the same capabilities that make legitimate chat products and image generators useful — persona consistency, fast iteration, low per-unit cost — are exactly what make these scams cheaper to run at volume.
What this means for platforms and builders
Teams building chat products, gaming platforms, or anything minors are likely to use should treat this as a design problem, not just a policy one:
- Age-appropriate friction on unsolicited job/payment offers inside gaming and chat platforms, especially ones that route users off-platform quickly.
- Detection tuned for grooming conversation patterns (rapid escalation toward secrecy, requests to move to unmonitored channels), not just keyword filters.
- Clear, low-friction reporting flows for minors specifically, since teenagers are less likely to escalate to a parent or platform support on their own.
- Provenance signals (watermarking, content credentials) on AI-generated images, so a deepfake extortion attempt is at least technically traceable back to the generation tool used.
None of these are silver bullets, and in our estimation, no single measure closes the gap on its own — the tactics described by Speka combine social engineering with generative tooling, so the response has to combine platform design with media literacy education rather than relying on automated detection alone.
AiiN's takeaway
The headline risk isn't that AI invented a new category of harm against teenagers — grooming, extortion, and fake job offers all predate it. The risk is that generative tools have removed the skill and time constraints that used to cap how many targets one bad actor could pursue at once. For product teams working on anything minors touch — games, chat, social — that's a reason to treat abuse-resistant design as a launch requirement, not a post-incident patch.