Overconfident consumers face higher fraud-loss risk when using AI financial advice

Consumers are increasingly using AI-powered financial tools for budgeting, investing, financial planning and fraud detection, while fraud losses are rising in scale and complexity. A new study published in the International Journal of Financial Studies reveals that AI tools may help some consumers navigate financial fraud risk, but overconfident users remain a weak spot as they are less likely to question bad advice, fake offers or deceptive financial claims.

Using data from the 2024 National Financial Capability Study (NFCS), a nationally representative sample of U.S. adults, the research "AI for Financial Advice, Fraud Loss, and the Moderating Effect of Financial Knowledge Miscalibration" examined whether willingness to use AI for financial advice is linked to financial loss from fraud, and whether that relationship changes when people misjudge the gap between what they think they know and what they actually know.

AI advice users showed lower fraud-loss odds, but the pattern is not simple

Scammers are also using digital technologies to make schemes more convincing, including imposter scams, investment scams, email scams and other forms of deception. In view of these developments, the researchers asked whether consumers who are open to AI financial advice face a different risk of losing money to fraud.

The findings show that willingness to use AI for financial advice was associated with lower odds of reporting financial loss due to fraud across all three regression models, suggesting that openness to AI may overlap with traits that protect consumers, such as greater comfort with digital tools, stronger online awareness or better use of technology-enabled safeguards. AI systems can also help identify irregular activity, detect suspicious patterns and support fraud prevention in financial services.

The researchers measured stated willingness to use AI for financial advice, not actual use of specific AI tools. This matters because a person may be open to AI without regularly using it, and willingness may reflect broader attitudes toward technology rather than direct protection from fraud. Willingness is a behavioral signal tied to digital financial engagement, trust in algorithmic systems and openness to technology-mediated decisions.

The study also narrows the fraud question to financial loss. For the unversed, fraud exposure and fraud loss are not the same. For example, a consumer may be targeted by a scam but avoid losing money. The researchers focused on people who reported being targeted by fraud and then examined whether they lost money as a result. 

The analysis included controls for income, employment status, marital status, gender and risk tolerance. The researchers also examined objective financial knowledge, subjective financial knowledge and financial knowledge miscalibration. Objective knowledge measured actual performance on financial questions. Subjective knowledge captured how people rated their own financial understanding. Miscalibration captured the gap between the two.

AI openness by itself was linked to lower fraud-loss odds, but the effect changed when confidence and knowledge were separated. Consumers who accurately understand financial concepts appear better positioned to spot risks. On the other hand, consumers who feel knowledgeable without matching objective knowledge may be more exposed, particularly when they are also willing to rely on AI for financial guidance.

Overconfidence changes the fraud-risk equation

Financial knowledge miscalibration, or overconfidence, happens when consumers rate their financial knowledge higher than their actual performance supports. The researchers found that overconfidence strengthened the association between willingness to use AI for financial advice and the likelihood of losing money to fraud.

Many fraud schemes exploit confidence as much as ignorance. A person who knows little may hesitate, ask questions or seek help whereas a person who thinks they know enough may move faster, verify less and trust their own judgment even when a situation is risky. In AI-mediated financial settings, this confidence can become more dangerous if users accept AI-generated information or digital recommendations without enough scrutiny.

Objective financial knowledge was linked to lower fraud-loss odds. Simply put, people with stronger actual financial knowledge were less likely to report losing money to fraud. However, objective knowledge did not significantly change the relationship between AI willingness and fraud loss. This means actual knowledge was protective on its own, but it did not meaningfully alter the AI-fraud relationship in the tested model.

Higher self-rated financial knowledge was associated with lower odds of fraud loss overall, but it also interacted significantly with willingness to use AI. The study found that fraud-loss probability rose at higher levels of subjective financial knowledge among respondents willing to use AI for financial advice. Confidence may help when it reflects real skill, but it can increase risk when it outruns actual knowledge.

The analysis utilizes Routine Activity Theory and Bounded Rationality.

• Routine Activity Theory suggests that routine behaviors can increase exposure to offenders when a person becomes a suitable target without enough protection. In financial life, digital engagement can create more points of contact with financial products, platforms and potential scams.
• Bounded Rationality explains why consumers do not always make fully informed decisions. Financial decisions are complex, and people often rely on shortcuts when information is incomplete or hard to process.

The theories help explain why overconfident AI-oriented consumers may be vulnerable. Their digital engagement may expose them to more financial offers or tools, while their confidence may reduce careful verification. They may believe they can tell reliable advice from deceptive claims, even when fraudsters use professional language, convincing interfaces or technology-assisted tactics.

The study also found other fraud-loss patterns. Higher risk tolerance was associated with greater odds of losing money to fraud. Lower-income respondents faced higher odds of fraud loss relative to the reference income group, while several higher-income groups had lower odds. Separated, divorced and widowed respondents showed higher odds of fraud loss than married respondents in the models, and retired respondents had lower odds than full-time employees. Males also showed higher odds of fraud loss in two models.

The findings suggest that fraud prevention cannot rely only on broad warnings. Risk varies by financial resources, household circumstances, risk tolerance, knowledge and confidence. The study adds AI willingness and knowledge calibration to that list, showing that the next stage of fraud prevention must account for how consumers interact with digital financial advice.

Why fraud policy must target overconfidence, not just ignorance

The willingness to use AI was associated with lower odds of fraud loss, which suggests AI tools may be useful when consumers have the skills and judgment to use them carefully. The sharper warning is that AI financial tools should not assume confidence equals competence.

The findings point to a need for calibration training for financial educators. Traditional financial education often focuses on improving knowledge about interest rates, inflation, debt, risk and diversification. The study suggests that education should also help consumers test how accurate their self-assessments are. People, particularly consumers who actively use digital financial tools, need to know not only financial facts, but also where their understanding is weak.

Fraud-prevention campaigns should also shift their focus from uninformed consumers to a high-risk group: people who are willing to use AI financial advice and believe they already have enough understanding of finance. For these consumers, the most effective message may be less about basic awareness and more about slowing down decisions, checking sources and verifying recommendations before making any transaction.

Financial professionals and consumer-protection groups can use this finding to design more targeted outreach. Consumers who overestimate their knowledge may need prompts that challenge quick decisions. Instead of discouraging AI use, campaigns could encourage practical safeguards like pausing before acting on unsolicited offers, verifying the identity of advisers or platforms, checking whether AI-generated recommendations come from a trusted source, and confirming financial claims through independent channels.

The findings prompt policymakers to rethink how AI financial advice platforms are designed, monitored and regulated. If AI tools are becoming part of consumer financial decision-making, platforms may need built-in protections that interrupt low-effort choices at risky moments. These could include plain-language risk alerts, confirmation steps before high-stakes transactions, fraud-awareness prompts and warnings when users appear to be acting on unsolicited advice or unusually risky recommendations.

The findings also call for stronger protections for lower-income consumers. Respondents earning below $50,000 had higher odds of fraud loss in the models. For households with fewer financial buffers, even smaller losses can have severe consequences. Lower-cost financial counseling, digital literacy programs and safe access to vetted AI financial tools could help reduce harm in communities where recovery from fraud is harder.

Last but not least, the findings leave important questions for future research. Because the data are cross-sectional, the results show associations rather than causation. The measure captures willingness to use AI, not actual AI use. The fraud measure is broad and does not separate online scams from offline fraud or identify specific scam types. Future studies will need to examine real-world AI use, fraud type, digital literacy and prior fraud exposure to determine where AI protects consumers and where it may create new vulnerabilities.