Fraud has always depended on pressure, imitation, and timing. What changed is the polish. In 2024, consumers reported losing $12.5 billion to fraud, and security agencies now warn that generative AI is helping criminals produce more convincing messages, cloned voices, fake documents, and highly personalized lures at greater speed. The result is not always a wildly obvious con. Increasingly, it is something that feels just plausible enough to trust for a few dangerous minutes.
The warning signs still exist. They are simply easier to overlook when the grammar is clean, the voice sounds familiar, and the story arrives wrapped in urgency. These 19 red flags capture the moments when suspicion should rise, even when the scam itself looks polished.
The opener that already knows a little too much
One of the easiest red flags to underestimate is a message that arrives with just enough personal detail to feel legitimate. It may mention a workplace, a recent purchase, a family role, or a hobby pulled from public posts, data leaks, or older online traces. That small hit of recognition lowers defenses fast because it makes the contact feel informed rather than random. In the AI era, that first impression matters even more because scammers can stitch together scraps of public information into a message that sounds tailored instead of mass-produced.
Research has shown why that matters. A late-2024 study on automated spear phishing found AI-generated personalized emails performed on par with expert human-crafted ones, while the model’s background profiles were accurate and useful in most cases. That helps explain why a simple text from a “recruiter,” “bank specialist,” or “customer-care agent” can feel eerily well aimed. The problem is not that the scammer knows everything. It is that knowing a little is often enough to get the conversation started.
An emergency that leaves no time to think
Urgency remains one of the oldest signals in fraud, but AI makes it easier to produce at scale and in multiple formats at once. A text claims a package will be returned today. A caller says suspicious activity is unfolding right now. A notice warns that a payment, account, or legal issue must be fixed within minutes. The common thread is not the story itself. It is the shrinking of the decision window until reflection feels like a luxury. When thought is replaced by reaction, the hook sets faster.
That tactic works because everyday life is already crowded with genuine alerts. Delivery updates, bank texts, and account notices are normal. The FTC reported $470 million in losses from text-message scams in 2024, and many of the most common schemes leaned on exactly this kind of deadline pressure. AI adds cleaner wording, more realistic formatting, and more convincing follow-up replies, so the warning sign is no longer sloppy writing. It is the demand to act before independent verification has a chance to catch up.
A voice or face that looks real but behaves strangely
Many people still expect an AI scam to sound robotic or look obviously fake. That assumption is increasingly outdated. Consumer warnings from the FTC note that a scammer may need only a short audio clip to clone a loved one’s voice, while European law enforcement assessments now explicitly warn that AI-powered voice cloning and live video deepfakes are amplifying fraud, extortion, and identity theft. A familiar voice is no longer strong proof of identity. It is only one data point, and sometimes a dangerously misleading one.
What often gives these scams away is not the overall effect but the tiny mismatch inside it. A voice may be emotionally flat in the wrong places. A face may move naturally but avoid unscripted interaction. In a 2025 case reported by Reuters, Italian police investigated a scam in which fraudsters allegedly impersonated defense officials and used calls that appeared to come from government offices to pressure prominent targets for money. That is the modern pattern: the performance feels real enough at first glance, while the seams show only if someone pauses long enough to notice them.
Pressure to move the conversation to a different app
A surprising number of scams do not begin with the main deception. They begin with a transfer. The first message is often mild, professional, or even bland. Then comes the pivot: move to WhatsApp, Signal, Telegram, a private SMS thread, or another channel outside the place where the contact began. That shift matters because it breaks the context that might help a target notice inconsistencies. It also gives the scammer more control over pacing, tone, and record-keeping.
The FBI highlighted this pattern in a 2025 warning about malicious campaigns impersonating senior U.S. officials. In those cases, the impersonator often established quick rapport on a topic the target recognized and then tried to move the exchange to a secondary, encrypted messaging app almost immediately. That same move appears in fake recruiting, fake investment mentoring, and romance-adjacent fraud. When someone claiming to be legitimate seems unusually eager to leave the official platform, the destination matters less than the reason: scammers prefer places where the pressure can become more personal and less accountable.
The request to keep it secret
Secrecy is one of the clearest scam signals because legitimate help rarely depends on isolation. Fraud does. The scammer wants the target separated from the one thing most likely to break the spell: another person’s common sense. Family-emergency scams have used this for years, telling victims not to alert parents, spouses, or siblings. AI only sharpens the effect by making the emergency sound more specific, more emotional, and more believable in the opening moments.
Federal consumer guidance states this pattern plainly. The FTC warns that fake-emergency scammers may say the situation is urgent, insist the victim is the only one who can help, and tell them to keep it secret so no one else checks the story. The FCC has documented similar “grandparent” scams, including calls where the supposed relative begs the target not to tell anyone. That red flag is easy to miss because secrecy can feel intimate or protective. In practice, it is often the point where manipulation stops pretending to be assistance and becomes control.
A code sent to a phone that “just confirms identity”
Few scam moves look more routine than the request for a one-time code. The victim is often told it is just a security step, a harmless confirmation, or proof that the support agent is helping the right person. In reality, that code is often the final barrier protecting an account. Once it is handed over, the scammer may use it to reset credentials, enter a banking session, or seize control of an account the victim believed was being protected.
The FTC has tried to make this point unmistakable: anyone asking for an account verification code is a scammer. That includes callers pretending to be from a bank’s fraud department. The trick works because the request arrives in a plausible setting, often after the target has already been rattled by warnings of suspicious activity. AI makes the surrounding conversation smoother and more reassuring, but the core rule has not changed. A one-time code is meant to prove identity to a system, not to a stranger explaining why the system can supposedly be trusted.
A warning that money must be moved to stay safe
This is one of the most destructive red flags because it sounds like protection rather than theft. The scammer says an account is compromised, an investigation is underway, or funds must be shifted before criminals get there first. Sometimes the supposed helper is from a bank, sometimes law enforcement, sometimes a government agency, and sometimes a retirement or investment firm. The wording changes, but the mechanism is constant: fear creates motion, and motion moves money toward the scammer.
The FTC has been blunt on this point because the losses are so severe. No legitimate fraud department, regulator, or government office will tell someone to move money to “protect” it. Yet people continue to comply because the instruction often arrives after a convincing setup involving spoofed calls, fake case numbers, or professional-sounding explanations. AI strengthens the script, not the logic. Once a conversation reaches the point where safety supposedly requires a withdrawal, transfer, or relocation of funds, the most important fact is simple: the money is no longer being protected. It is being lined up for removal.
Payment instructions built around crypto, gift cards, gold, or Bitcoin ATMs
The payment method itself can be the red flag. Scammers prefer instruments that are fast, hard to reverse, or difficult to trace. That is why so many schemes end with cryptocurrency, wire transfers, gift cards, couriered cash, gold bars, or cash inserted into a Bitcoin ATM. These methods do not just move money. They reduce the victim’s chance of recovery, which is exactly why fraudsters keep steering people toward them.
The numbers show how costly that guidance can be. FTC data for 2024 showed the biggest scam losses came through bank transfers and payments, with cryptocurrency close behind. Separate FTC reporting found losses at Bitcoin ATMs rose dramatically, topping $65 million in just the first half of 2024, with a median reported loss of $10,000 during that period. In real life, these requests are often wrapped in polished explanations about security, speed, or compliance. But the moment a stranger insists that unusual payment rails are the only safe option, the disguise is already slipping.
A link or QR code offered as the fastest fix
Convenience is one of the most effective scam disguises. The message says there is a simple next step: tap the link, scan the QR code, confirm the charge, re-route the package, verify the account, or claim the refund. Because the task seems tiny, many people treat it as low risk. That is exactly the trap. A short action can open a much larger door, whether to credential theft, malware, or a fake site designed to collect payment details.
Both the FTC and the UK’s National Cyber Security Centre have warned about this pattern in newer forms, including malicious QR codes. The FTC has cautioned that QR codes on unexpected packages can lead to phishing sites or malware, while the NCSC notes that QR-related fraud often relies heavily on social engineering around a believable setting. What matters is not whether the code or link looks modern. It is whether the problem could have been verified another way. When the fastest route is also the least independent one, the risk has usually been engineered on purpose.
An address, domain, or caller ID that is almost right
Modern scams often win by being close enough. The email domain differs by one letter. The text thread looks official. The caller ID displays a familiar institution. The paid search result appears above the real one. None of those details guarantees authenticity, and criminals know most people are trained by daily life to accept surface familiarity as proof. In practice, “almost right” is one of the most common ways a scam gets over the line.
The FCC continues to warn consumers about caller ID spoofing, where callers deliberately falsify what appears on the screen. The FBI has also warned that criminals have used search-engine ads to impersonate legitimate employee self-service websites, while FTC guidance now advises people not to rely on top search results for a company’s contact details because scammers often buy those placements. That combination matters. The red flag is no longer just a typo in a suspicious email. It is the broader realization that search rank, display name, and caller ID can all be manipulated to create borrowed trust.
A refund, invoice, or subscription panic that starts the whole chain
One of the smartest scam openers is not a promise but a problem. An invoice appears for something never bought. A subscription seems about to renew for an absurd amount. A text says a refund is waiting, but only after a quick confirmation. These messages work because they trigger defensive action. People do not want a prize badly enough to ignore warning signs, but they do want to stop an unauthorized charge before it lands.
The FTC has documented repeated versions of this tactic, from fake invoices to bogus renewal notices and texts offering “refunds” for large-brand purchases. The typical sequence is familiar: the target is told to call a number, click a link, or let an agent “help” correct the transaction. From there, the scam often turns into credential theft, remote-access fraud, or direct payment demands. AI improves the props by generating cleaner invoices, more natural replies, and more believable support language. The underlying red flag remains the same: a surprise billing scare that channels the target toward the contact method chosen by the scammer.
A recruiter who hires too fast and asks for money or personal data
Job scams succeed because they exploit hope, urgency, and social pressure all at once. A message arrives unexpectedly with a flattering tone, promising remote work, fast pay, or immediate openings. The hiring process is strangely effortless. There may be no meaningful interview, little scrutiny, and quick encouragement to move forward before the opportunity disappears. That speed feels lucky when someone wants work. In reality, it often signals that the real product is not employment. It is the applicant’s money or identity.
The FTC has recently warned about fake recruiter texts and longer-running impersonation scams on LinkedIn and other platforms. Consumer guidance is consistent: honest employers do not ask people to pay for a job, front equipment costs, or deposit a check and send some of the money elsewhere. The FBI has also described cryptocurrency job scams that require victims to deposit their own funds to complete supposed tasks. AI makes fake recruiters sound more polished and personalized, but legitimate hiring still follows the same basic rule it always has: employers pay workers, not the other way around.
An online relationship that suddenly becomes a financial guide
A romantic connection turning into investment coaching is not a quirky modern storyline. It is one of the more expensive fraud pathways online. The target meets someone who seems attentive, patient, and unusually competent with money. The relationship builds emotional trust first, then pivots into financial trust. By the time the investment recommendation appears, it feels less like salesmanship and more like care. That emotional sequencing is what makes the red flag so easy to miss.
The FTC has warned directly that if someone met online offers to help with cryptocurrency investing, it is an investment scam. The same agency has also reported enormous romance-scam losses, with 2023 reports totaling $1.14 billion and median losses higher than for other forms of imposter fraud. The FBI’s cryptocurrency-fraud guidance describes how victims are persuaded to keep depositing more money into investments that are entirely fake. AI adds realism to the mentor figure, the trading interface, and the ongoing conversation, but the pattern remains old and brutal: intimacy is being used to escort money toward a fake opportunity.
Celebrity or expert endorsements that feel more vivid than real
A persuasive endorsement used to require an actual public figure, or at least a crude imitation. Now a scam ad can pair a familiar face, a convincing voice, and confident scriptwriting into something that feels like social proof. The result is powerful because people do not always remember the exact claim. They remember the feeling that someone recognizable seemed to validate it. That emotional residue is often enough to lower skepticism.
U.S. investor guidance now explicitly warns that AI-enabled fraud can involve deepfake video and audio, including content designed to impersonate trusted people and steer investors into bad decisions. The FTC has also warned that scammers are using doctored video and audio to fabricate celebrity and influencer endorsements for products and money-making schemes. This is why “but the video looked real” is becoming less useful as a defense. In many cases, realism is the bait. The real red flag is an endorsement that asks for trust before there is any independent proof that the person, offer, or platform is genuine.
Refusal to verify through ordinary, low-tech steps
Real institutions usually do not mind being verified. Scammers hate it. A fake recruiter resists a callback through the company’s main line. A supposed bank agent discourages use of the official app. A love interest always has a reason a live meeting cannot happen. A “support” contact wants the interaction to stay inside the message thread that started the panic. The problem is not that verification feels inconvenient. It is that scammers often turn inconvenience into an argument against verification itself.
That resistance is a major warning sign because low-tech checks still defeat high-tech deception surprisingly often. The FTC advises people to verify a story using a phone number, website, or app they know is real rather than contact details supplied in the suspicious message. Romance-scam guidance similarly warns that scammers often cannot meet in person and recommends outside checks such as reverse-image searches. AI can generate a smooth persona, but it cannot make a fake relationship withstand real-world scrutiny forever. When someone opposes the easiest independent test, the excuse is rarely the truth.
Requests for remote access, screen sharing, or a “security tool”
A scam often becomes far more dangerous the moment a stranger is allowed onto a device. Tech-support schemes have used this trick for years, but AI can now make the approach sound calmer, more technical, and more legitimate. The script may involve a charge dispute, malware alert, refund correction, or account-security problem. Once the victim is persuaded to install software, share a screen, or hand over device control, the scammer’s options multiply quickly.
The FTC’s consumer guidance on tech-support scams remains strikingly relevant because the structure has not changed. Victims are often pushed toward remote access, spoofed websites, and fake troubleshooting steps that culminate in stolen financial information or direct account access. The agency has also warned that impostors, including those pretending to be from the FTC itself, have used promises of help or refunds to get onto victims’ computers. This red flag matters because remote access feels procedural, even routine. In reality, it can be the precise moment when suspicion should become an immediate no.
New payment instructions that skip normal approval steps
In business settings, the danger signal is often procedural rather than emotional. A vendor’s banking details suddenly change. An executive needs an urgent transfer outside the usual process. An invoice must be paid now, despite bypassing standard checks. Because the request arrives inside a familiar workflow, it can seem like an exception rather than a threat. That is what makes business email compromise and related fraud so persistent: the scam borrows trust from existing operations.
The FBI has long described BEC as one of the most financially damaging online crimes, and the bureau’s 2025 reporting showed it remained among the biggest loss categories, at roughly $3 billion. The method is straightforward and devastating: criminals make a message appear to come from a known source and use that legitimacy to redirect funds. AI increases the danger by improving tone, formatting, and plausibility across email, voice, and chat. When payment instructions arrive with unusual urgency and weakened process controls, that is not efficiency. It is often the attack surface opening.
Tiny inconsistencies that get brushed aside because the overall story feels polished
A polished scam can survive on the strength of its atmosphere. The victim notices one odd phrase, one strange pause, one slightly vague answer, one mismatched detail, but dismisses it because everything else seems coherent. That is a dangerous instinct. The more convincing a scam feels overall, the more tempting it becomes to excuse the small fracture instead of testing it. AI-generated fraud is especially good at this: broad plausibility with local mistakes hidden inside confidence.
Consumer fraud guidance has long warned people to slow down and look for inconsistent answers, especially in romance and impersonation scams. That advice matters even more now because AI can smooth out the bigger rough edges that once gave scammers away quickly. In other words, a scam no longer has to be flawless. It only has to be fluent. The correct response to a small inconsistency is not to explain it away. It is to treat it as a chance to probe. Often the smallest crack is the first honest thing in the entire interaction.
Isolation inside a chat group, feed, or conversation that drowns out doubt
Some of the most effective scams do not just create a lie. They create an environment around the lie. A victim is added to a busy group chat, surrounded by people who seem enthusiastic, informed, and successful. Social posts, testimonials, and AI-generated “tips” reinforce the same message from different angles. Suspicion fades because the target is no longer evaluating a single claim. They are absorbing a miniature world designed to make the claim feel normal.
Recent enforcement and consumer data show how costly that environment can be. The FTC reported that losses to scams originating on social media reached $2.1 billion in 2025, while the SEC described a 2025 scheme in which fraudsters used social-media ads, group chats, and supposed AI-generated investment tips to lure people onto fake crypto platforms. That combination is potent because it replaces independent judgment with ambient confidence. When every surrounding voice appears to agree, the real red flag is not the agreement. It is the fact that the agreement arrived prepackaged.
19 Things Canadians Don’t Realize the CRA Can See About Their Online Income
Earning money online feels simple and informal for many Canadians. Freelancing, selling products, and digital services often start as side projects. The problem appears at tax time. Many people underestimate how much information the CRA can access. Online platforms, banks, and payment processors create detailed records automatically. These records do not disappear once money hits an account. Small gaps in reporting add up quickly.
Here are 19 things Canadians don’t realize the CRA can see about their online income.
