AI is getting better at sounding like someone you know, understanding what makes you tick, and finding just the right words to get you to do something. That becomes particularly relevant when holiday distractions, family emergencies and the pressure to get things done can make anyone less cautious than usual.
The good news is that beating these tricks doesn’t require a degree in computer science, just a better understanding of how manipulation works and a few habits that are surprisingly difficult for scammers to get around.
📊 Getting Better Poll
Share this post and invite others to share what informed their answer or shaped their perspective.
▶️ If you would like a wider lens on how fear and sensational stories shape public trust, check out our award-winning documentary “Trust Me”.
🧠 When AI Knows Exactly What to Say to Change Your Mind
Johannes C. Zeller explores how conversational AI could take social engineering beyond traditional phishing. Rather than sending thousands of identical messages, AI can personalize conversations, adapt its arguments and use information people have voluntarily shared to make persuasion more effective. The interesting distinction is that manipulation doesn’t always look like a scam. Sometimes it looks like a helpful conversation with something that seems to understand you.
Zeller draws on established psychological research to suggest six ways to become harder to manipulate. These include recognizing that nobody is immune, stepping away from persuasive conversations, limiting personal information shared with AI, maintaining emotional distance and checking what a system is designed to achieve. Importantly, the author acknowledges that these are reasonable precautions, not yet proven defences against AI manipulation. The longer-term lesson is that critical thinking is becoming a digital survival skill, not just a classroom exercise.
The Better Take:
Being persuasive isn’t the same as being right. Enjoy the convenience of AI, but keep control of your decisions. Ask who created this message, what are they trying to get me to do, what evidence supports its advice, consider what it might be leaving out, and check important claims independently. A little healthy skepticism makes technology more useful without turning every conversation into an interrogation.
👨👩👧👦 That Familiar Voice Might Not Be Who You Think
The Federal Trade Commission is drawing attention to a particularly personal form of AI-enabled fraud: scammers impersonating relatives in distress. The scenario is unsettlingly ordinary. Someone receives an urgent call from a grandchild who claims to need money for bail, car repairs or another emergency. The caller sounds familiar, asks for immediate help and insists that the conversation remain secret. Scammers can use cloned voices, fake videos and personal details collected from social media to make the story more convincing.
The important distinction is that a convincing voice or video is no longer reliable proof of identity. The FTC recommends slowing down, ending the call and contacting the supposed family member through a number already saved in your phone. It also suggests involving another trusted person before sending money. This is particularly relevant around family gatherings and holiday travel, when unexpected emergencies can sound plausible and people may be harder to reach. The underlying scam is old. The impersonation technology is getting an upgrade.
The Better Take:
A family verification routine can be more useful than trying to identify a perfect voice clone. Agree on a family code word, keep trusted numbers accessible and establish a rule that emergency money requests must be independently confirmed. These small preparations protect relationships as well as bank balances, without requiring anyone to become a technology expert.
📊 The Numbers Behind the Deepfake Problem
AI impersonation is no longer just a theoretical cybersecurity scenario. A Gartner survey of 297 senior cybersecurity leaders, conducted between March and May 2026, found that 41% had reported at least one social engineering incident involving a deepfake during an employee audio call in the preceding 12 months. Another 36% reported an incident involving a video call. For context, 79% reported email phishing, spear-phishing or business email compromise incidents. These figures describe reported organizational experiences, not the percentage of individuals who have been scammed.
The bigger development is how AI changes the economics of deception. Attackers can combine personal information, convincing synthetic media and tailored messages across different communication channels. Gartner argues that organizations should move beyond simply teaching employees to spot suspicious content and instead make identity verification standard practice. That principle matters outside the workplace, too. As familiar visual and audio clues become less dependable, checking who is making a request becomes more valuable than judging how convincing they sound.
The Better Take:
The numbers are a reason to improve verification, not to distrust every voice or video. Make a habit of confirming unusual requests through a separate, trusted channel. At work or at home, a simple identity check can accomplish what increasingly sophisticated fake detection tools cannot reliably guarantee.
📚 Getting Better Recommendations
Teen safety: 6 tips for AI and social media infographic, created by the News Literacy Project in partnership with the American Psychological Association, is a resource for parents, families and community members. It shares six safety tips to guide teens as they use AI tools and social media platforms.
