In an era where scams increasingly move online, a new study has delivered a sobering finding: artificial intelligence can be better than humans at building the trust needed to pull off devastating romance-based fraud. The research, published in the journal Communications Psychology, reveals that AI chatbots outperformed human operators in establishing interpersonal closeness during emotionally engaging interactions—but only when the AI was labeled as human to the unsuspecting target.

This isn't just a theoretical exercise. The study specifically simulated "pig butchering," a growing global scam in which fraudsters spend weeks or months cultivating a romantic or friendly relationship with victims before luring them into fake cryptocurrency investments. The financial toll is immense, with tens of billions of dollars stolen annually worldwide, often leaving victims with six-figure losses.

Inside the Study: AI vs. Human Scammers

Researchers from four universities—Amrita Vishwa Vidyapeetham in India, Foscari University of Venice in Italy, the University of Melbourne in Australia, and Ben Gurion University of the Negev in Israel—carried out a broad experiment to test whether generative AI chatbots could replace human scammers in the crucial trust-building phase. They pitted AI chatbots directly against human confederates in simulated conversations designed to mimic the early stages of a pig-butchering scam, where the goal is to forge an emotional bond before any financial request is made.

The results were striking. In the relationship-establishing stages—which in real-world scams often stretch over months—the AI chatbots performed remarkably effectively, at times exceeding the ability of human scammers to build rapport and interpersonal closeness. As Ars Technica reported, the notion that scammers can use AI to "sharpen their deceptions, polish their language, and lubricate their banter with victims" is no longer hypothetical.

The 'Labeled as Human' Catch

One of the most nuanced findings, highlighted in the study's title, is that AI's advantage disappeared when participants were explicitly told they were interacting with an AI. The paper states: "AI outperforms humans in establishing interpersonal closeness in emotionally engaging interactions, but only when labelled as human." This suggests that deception itself is a crucial component: victims are more likely to open up when they believe they are speaking to another person, even if the AI's conversational skills are superior.

"AI outperforms humans in establishing interpersonal closeness in emotionally engaging interactions, but only when labelled as human."

This finding has serious implications for fraud prevention. It means that simply warning users about AI-driven scams may not be enough; people must also verify the identity of who they're talking to online. The study underscores that AI can autonomously carry out the majority of a long con, from initial contact to the emotional bonding that precedes a financial ask, potentially at scale.

The Broader AI Scam Landscape

The study arrives amid a wave of concern about AI-powered fraud. As tech publications like Wired emphasized, AI scammers are being used to build trust in everything from romance scams to investment fraud. But the problem extends beyond text-based chatbots. Deepfake technology is making these scams even more believable, as MSN's coverage noted, with scammers using AI-generated voice and video to impersonate trusted figures. In 2026, the security landscape is expected to be dominated by hyper-personalized attacks that leverage data scraped from social media.

Bitdefender, a major cybersecurity firm, also points to social media platforms as the new "scam inbox." Fraudsters use AI to craft phishing messages that mimic friends or influencers, making it increasingly difficult for users to distinguish between genuine interactions and automated deception. The study's findings add empirical evidence to these warnings, showing that AI's ability to forge interpersonal closeness is not just theoretical but measurable in controlled conditions.

Links to Human Trafficking and Organized Crime

The research from Amrita University has also drawn attention to a darker connection: the link between AI scams and human trafficking. According to News Mobile, the study highlights that many pig-butchering operations are not just run by individuals but by organized crime syndicates that traffic victims to work as scammers in forced-labor compounds. AI could now reduce the need for human operators, potentially shifting the economics of these criminal enterprises. While the study didn't specifically measure trafficking, the implications are clear: if AI can replace the human talent needed to groom victims, criminal networks may become even more efficient and harder to dismantle.

Expert Reactions and Implications

Cybersecurity experts are divided on the severity of the threat. Some, like those who spoke to WIRED, argue that AI is a double-edged sword: it can also be used to detect and counter scams, analyzing patterns of fraudulent conversation in real time. Others warn that the asymmetry of advantage currently favors the attackers, particularly as AI models become more sophisticated and accessible.

For the general public, the takeaway is to be more cautious than ever in online relationships, especially when money is involved. Authorities and financial institutions are also under pressure to develop better detection systems and public-awareness campaigns. The study's authors suggest that future research should explore how AI can be "red-teamed"—tested by security experts to identify vulnerabilities—and how to build tools that can automatically flag AI-generated conversations.

Conclusion: A New Arms Race

This study marks a watershed moment in the evolution of online fraud. As AI chatbots become more adept at mimicking human empathy and trust, the line between human and machine deception is blurring. The findings from Amrita, Venice, Melbourne, and Ben-Gurion universities serve as both a warning and a call to action. For consumers, the old adage holds true: if something feels too good to be true, it probably is—especially if the person on the other side of the screen is an algorithm. For defenders, the next battle against cybercrime will be fought not only in code, but in the psychology of trust itself.