The Geek HR editorial team has been tracking a straightforward but unsettling shift in the online threat environment facing Croatian households: the signals that once made internet scams easy to spot have largely disappeared.

According to The New York Times, low-cost chatbots, image generators, and voice-cloning tools now allow criminals to produce polished, professional-looking fakes — pristine copy, convincing websites, and near-perfect identity replication — at a scale that was unimaginable only a few years ago. The old checklist is obsolete.

What has replaced it is a habit-based approach built around three pillars: independent verification, caution around payments, and strong authentication.

How Generative AI Removed the Tells That Protected People

For a long time, the signs of a scam were almost comically obvious. An email riddled with typos, a blurry product photo, a customer service voice that stumbled over basic phrases — these were reliable signals that something was wrong. That era is effectively over.

Generative AI tools have handed criminals capabilities that strip those signals away entirely. Chatbots produce flawless, natural-sounding prose in any language. Image generators create product photography and brand assets indistinguishable from legitimate commercial work.

Voice-cloning software captures the timbre and cadence of a real person’s voice from only a short audio sample. Each of these tools is cheap, widely accessible, and requires no technical background to operate. Together they close off the shortcuts that ordinary users relied on to filter out fraud.

Because building convincing websites and digital avatars now requires almost no effort or skill, security experts say the entire mental model for recognizing fraud has to change. Readers cannot trust surface appearance — tone, visual polish, or a familiar-sounding voice — the way they once could.

FBI Numbers Put the Scale of the Problem in Concrete Terms

The financial damage is not abstract. The FBI reported that cybercriminals defrauded Americans of nearly $21 billion in 2025, a figure that captures the full weight of the current environment.

Of that total, roughly $893 million was linked specifically to AI-enabled fraud. That figure represents the documented, attributable share — the portion where investigators could draw a direct line from a criminal tool to a victim’s loss. The true AI-linked total is almost certainly larger, since many cases do not surface or are not categorized with enough precision to be counted.

Neither number covers Croatian or European losses, but the mechanics of AI-powered scams are not geographically limited. The tools, the techniques, and the impersonation vectors cross borders without friction.

Fake Storefronts, Celebrity Impersonators, and Relatives in Distress

Three scam types in particular have grown in scope and sophistication as AI tools have matured.

The first is the look-alike storefront. Criminals build websites that closely mimic the pages of real brands, complete with logos, product photography, and customer-service copy that passes a quick visual check. The New York Times writer encountered one firsthand: a TikTok advertisement offered Hoka sneakers at 80 percent off, and clicking through loaded a site that appeared to be a legitimate brand clearance outlet.

The deception was good enough to nearly succeed. Hoka itself published a warning about a surge of fake web stores impersonating its brand, and Reddit users had already reported losing money to the same fraudulent site before the article appeared.

The second type involves criminals posing as celebrities, using AI-generated audio, video, or text to manufacture endorsements and build false credibility around financial schemes or product promotions.

The third — and in some ways the most emotionally disorienting — is the family impersonation scam. Voice-cloning tools make it possible for a criminal to reproduce a specific person’s voice convincingly enough to call a relative and ask for emergency money.

A parent hears what sounds exactly like a child’s voice describing an urgent crisis. The social and emotional pressure in that moment is intense, and the deception is hard to detect in real time.

Verification, Payment Caution, and Authentication as the New Standard

Given that AI makes websites, voices, and written communication simple to fake, the protective habits that matter most are those that operate independently of surface appearance.

Independent verification is the first principle. When a message, call, or website prompts any kind of action — especially financial — checking through a separate, known-good channel before responding is the approach security experts endorse.

A voice that sounds like a family member can be confirmed with a callback to a stored number. A storefront that looks legitimate can be cross-checked against a brand’s official domain found through a direct search rather than a link.

Caution around payment methods forms the second layer. Certain payment types offer meaningful consumer protections; others, once sent, are essentially irreversible. The payment mechanism a requestor insists upon is itself a signal worth scrutinizing.

Strong authentication, particularly for financial and email accounts, addresses the credential-theft vector that often underpins these schemes. When an account is protected by something beyond a password alone, the impact of a successful phishing attempt is contained.

Staying Current as the Threat Keeps Shifting

The Geek HR editorial team notes that the hardest part of this threat environment is its pace. AI capabilities do not stay static: each new development in voice cloning, image generation, or chatbot fluency opens a fresh impersonation vector, and any fixed list of warning signs ages out quickly. No single article, including this one, can serve as a permanent guide.

The team follows Geek Crotia precisely because the AI-driven tactics that power these scams continue to mutate, and readers need coverage that tracks those shifts as they happen rather than a snapshot that reflects last year’s threat landscape.

The underlying condition driving all of this remains unchanged: the tools are inexpensive, accessible, and improving. For anyone navigating the current web, the three habits — verify independently, scrutinize payment requests, and protect accounts with strong authentication — are not a temporary fix. They are the new baseline.