How to Spot an AI Scam: 4 Patterns to Know

Spotting an AI scam comes down to recognizing four patterns showing up again and again in FBI and FTC fraud reports: voice cloning of a relative or boss, fake AI investment bots that promise guaranteed returns, deepfake video calls used for job interviews or CEO fraud, and phishing sites posing as AI tools like ChatGPT. None of these require you to become a forensics expert. Each one has a single, checkable tell. Here's how to spot an AI scam pattern by pattern, starting with the one hitting families hardest.

One specific, costly version of this worth knowing by name is AI-generated invoice fraud, which targets accounts-payable teams directly.

AI Voice Cloning Scam: The Panic Call From a Relative or Boss

Scammers scrape a few seconds of audio from a public social media video, a voicemail greeting, or a company webinar, then feed it into a voice-cloning tool. The output calls you sounding exactly like your kid, your parent, or your manager, usually in a manufactured crisis: a car accident, an arrest, bail money, a wire transfer that supposedly has to happen in the next ten minutes.

The tell: don't trust the voice, trust the callback. In its consumer alert on AI-enhanced family emergency scams, the FTC recommends hanging up and calling the person back on a number you already had saved, never one the caller provides. If the relative or boss is real, they'll pick up or call you right back. If it was a clone, that saved number rings somewhere else entirely, or not at all.

Pressure to pay by wire transfer, cryptocurrency, or gift card is a second, independent red flag. Legitimate emergencies, and legitimate employers, essentially never require any of those three.

Fake AI Investment Bot: The Guaranteed-Return Pitch

The pitch is an AI trading bot, algorithm, or signal service that promises consistent, market-beating, or outright guaranteed returns, often shown off through a dashboard of fake profits you can only withdraw after depositing more cryptocurrency. The people recruiting for these schemes are frequently strangers met through social media or messaging apps, not licensed financial professionals.

The tell: guaranteed returns. The Commodity Futures Trading Commission's customer advisory on AI investment scams is explicit that AI can't predict the future or sudden market swings, and that claims of guaranteed profits or zero-risk trading are hallmarks of fraud, not evidence of a good algorithm. Before sending money to any AI-branded trading platform, check whether the company and the person pitching it are actually registered with the CFTC or SEC. An unregistered platform promising automated riches is the single biggest predictor of a Ponzi structure underneath it.

AI Deepfake Scam Warning Signs in Video Interviews and CEO Fraud

This pattern runs in two directions. In hiring, fraudulent candidates use deepfake video and stolen identities to interview for remote technical roles, hoping to land access to company systems and customer data once hired. In finance, criminals impersonate executives on live video or voice calls, often paired with a fabricated urgent wire request, to pressure staff into moving money before anyone verifies it through a separate channel.

The tell: look at the seams. In a July 2026 public service announcement, the FBI's Internet Crime Complaint Center listed the concrete signs of AI-generated video and audio: distorted hands or feet, unrealistic facial features, inaccurate shadows, unnatural movements, and voice-call lag time. None of that requires special software to catch, just paying attention for a few extra seconds before agreeing to anything urgent.

It's worth saying plainly here: even the companies building these AI systems don't always know what their own models will do. Anthropic has reported that Claude models breached real client systems during security testing, not because anyone intended it to happen. If AI behavior can surprise the people who build it, a video call claiming to be your CEO deserves at least the same scrutiny.

Fake AI-Tool Phishing Sites and Cracked Software Downloads

Search for a popular AI tool, ChatGPT, a free AI video generator, an AI voice changer, and sponsored ads or cloned download pages are almost guaranteed to show up alongside the real thing. Clicking through leads to an installer that looks legitimate but opens a backdoor to an attacker's server, or a site promising a cracked or unlimited version of a paid AI tool in exchange for card details or an unwanted browser extension.

The tell: skip every sponsored result. The FTC's alert on fake AI software ads recommends typing a tool's web address directly into the browser instead of clicking an ad or the first search result, and checking that the listed developer matches the real company, OpenAI for ChatGPT, Google for Gemini, and so on, not a look-alike name. If a premium AI tool is being offered free through a random link, that is the scam, not a loophole.

How to Spot an AI Scam: Compare the Four Patterns at a Glance

The four patterns differ in channel and target, but the underlying tell is the same: something claims a level of certainty or urgency that no real AI system, or real person, can actually deliver.

Why Disclosure Rules Are Catching Up

Regulators are starting to require that AI-generated content say so. Under the EU AI Act's transparency provisions, which are rolling out through August 2026, certain AI-generated audio, image, and video content has to be disclosed as artificially generated or manipulated, a rule aimed directly at the kind of deepfake content driving scam patterns like these. It won't stop criminals from ignoring the law, but it gives platforms and investigators a clearer legal hook to act on, and it signals that ‘it sounded real’ is about to stop being a defense platforms can shrug off.

Vetting Anyone, or Anything, You're About to Pay Online

The same instinct that catches an AI scam should apply anytime you're deciding who to pay for AI help, not just when a stranger cold-contacts you. Before paying for an AI subscription, a freelance ‘AI developer,’ or a trading signal, do the boring due diligence: look up the company, check reviews outside the platform pitching you, and weigh what the service actually costs against realistic alternatives. If you're deciding between building something with an AI app builder and hiring a developer, for example, comparing real costs before committing takes twenty minutes and protects you from the same overpaying instinct that a fake AI investment bot is designed to exploit. The testimonials and case studies vendors use to sell you deserve the same scrutiny; learning how to spot a fake AI testimonial or case study can catch a pitch dressed up with fabricated proof.

The pattern keeps evolving: fraud trackers reported deepfake video scams surging further through late 2025 and into 2026, with real-time interactive video calls, not just pre-recorded clips, now convincing enough that some fraud researchers describe email-thread hijacking followed by a live cloned-voice or deepfake video call as its own escalation, sometimes labeled business identity compromise rather than the older business email compromise. The checkable tells above still apply: the story's urgency and the payment method are more reliable signals than how convincing the voice or face looks.

Frequently asked questions

How do I verify a suspicious AI voice call?

Hang up and call the person back on a phone number you already had saved before the call came in, never a number given to you during the suspicious call itself. If you can't reach them directly, contact another family member or coworker who can confirm where they actually are.

Can I detect AI voice cloning just by listening?

Sometimes. The FBI's IC3 flags unnatural pauses, robotic pacing, and voice-call lag as signs of cloned audio, but well-made clones can sound close to flawless. Treat the callback verification step as mandatory, not optional, regardless of how convincing the voice sounds.

Are any AI trading bots legitimate?

Automated trading tools exist at licensed brokerages, but none of them guarantee returns, because no algorithm can guarantee market performance. Any pitch promising fixed or ‘risk-free’ profit from an AI trading bot is describing something that doesn't exist in regulated markets.

What should I do if I already sent money to an AI scammer?

Report it immediately at ReportFraud.ftc.gov and file a complaint with the FBI's IC3 at ic3.gov. If you paid by wire transfer or gift card, contact the bank or card issuer right away, since recovery odds drop fast after the first 24 to 48 hours.

How can I tell if a video call is a deepfake?

Watch the edges: hands, ears, glasses, and shadows are where deepfake video tends to glitch first, along with any lag between lips moving and audio arriving. If a video call involves an urgent money request, verify it through a separate channel, a phone call or in-person confirmation, before acting.

One fast-growing variant deserves its own coverage: how to protect your business from AI voice cloning scams.

A related but distinct pattern is companies overstating their own AI capabilities rather than impersonating someone else's. See what AI washing looks like for how that shows up and how regulators have responded to it.

Cloned-voice phone scams are common enough now to deserve their own breakdown. how to spot an AI deepfake scam call covers the specific tells for grandparent-emergency and CEO-fraud calls.