How to Spot AI Scams and Deepfakes Before You Pay

AI has made scams cheaper to run and harder to spot. Criminals now use cloned voices, fake videos of celebrities, realistic photos and fluent, error-free messages to trick people into sending money or sharing passwords. The good news is that the warning signs are still recognisable once you know them. This guide explains the most common AI scams and deepfakes in 2026 and gives you a simple routine for checking anything suspicious before you pay.

Why AI makes scams more convincing

Old scam messages were often easy to spot because of poor spelling, strange grammar or generic greetings. Generative AI removes many of those clues. A scammer can produce a polished message in any language, generate a profile photo of a person who does not exist, or clone a voice from a short clip posted online. Video deepfakes still have flaws, but they are improving and can be convincing on a small phone screen.

What has not changed is the goal. Nearly every scam tries to make you act quickly, emotionally and secretly, usually by moving money or handing over access. That is what you should watch for.

The most common AI scams

Voice-cloning emergency calls

You receive a call or voice note that sounds like a family member in trouble: an accident, an arrest, a stolen phone. They need money urgently and ask you not to tell anyone. The voice may be generated from videos the person posted on social media.

Deepfake celebrity investment adverts

Videos appear to show well-known business leaders, presenters or politicians recommending a trading platform, often described as AI-powered or "quantum". These are fake. Our guide What Is Quantum AI? Separating Facts From Hype explains this scam in more detail.

Fake bosses and business email compromise

An email, message or even video call appears to come from a manager or supplier asking for an urgent payment or a change of bank details. AI helps scammers copy writing style and, in some reported cases, faces and voices on calls.

Romance and friendship scams

Profiles built with AI-generated photos and chat replies build trust over weeks before asking for money, gifts or "investment" help.

Fake shops and customer support

AI makes it easy to create professional-looking shop websites, product photos and reviews. Fake support accounts on social media reply to complaints and direct victims to phishing pages.

Job and loan offers

Messages offering remote jobs, quick loans or grants ask for an upfront "processing fee" or personal documents that are later used for identity fraud.

How to spot a deepfake

No single test is reliable, but these clues help:

• Audio: flat emotion, odd pauses, unusual pronunciation of names, or a refusal to switch to a video call.
• Video: mouth movements slightly out of sync with speech, blurring around the face or hairline, unnatural blinking, or lighting that does not match the background.
• Images: distorted hands, jewellery or text, mismatched earrings, or backgrounds that do not make sense.
• Context: a famous person promoting a get-rich-quick scheme, or a video that only exists on one unknown account.

Some legitimate publishers now attach Content Credentials to their media using the open C2PA standard, which records how a file was made and edited. A missing label does not prove something is fake, but a verified credential from a trusted source is reassuring. For more on labels, see AI Transparency: How to Label AI-Generated Content.

A five-step check before you pay

1. Pause. Urgency is the scammer's main tool. A genuine person or organisation can wait ten minutes.
2. Verify through a different channel. Hang up and call the person back on a number you already have. Contact your bank through its official app. Do not use contact details given in the suspicious message.
3. Ask a question only they would know. Many families now agree a private code word for emergencies.
4. Check the payment request. Requests for gift cards, cryptocurrency, payment to a personal account or a "new" bank account are major red flags.
5. Talk to someone. Scammers ask for secrecy. Telling a friend or colleague often breaks the spell.

Protect yourself in advance

• Limit how much of your voice and video is public, especially on open social media accounts.
• Turn on two-step verification for email, banking and social media.
• Never share one-time passwords, PINs or recovery codes, whoever asks.
• Set up a family code word and agree that money requests will always be confirmed by a call.
• In a business, require two people to approve payments and changes to supplier bank details.

The UK National Cyber Security Centre's top tips cover passwords, updates and account security in plain language.

What to do if you have been scammed

Act quickly. Contact your bank immediately through its official channels and ask it to stop or recall the payment. Change passwords on any accounts you shared, starting with email. Report the scam to the police and to the platform where it happened. In the United States, the FBI's Internet Crime Complaint Center accepts reports; other countries have similar services. Keep screenshots, phone numbers and transaction references. Be wary of "recovery agents" who contact you offering to get your money back for a fee, which is a common follow-up scam.

Using AI to check a suspicious message

AI can also help you. Paste the text of a suspicious message, with your personal details removed, into an assistant such as TextPixAI AI Chat and ask: "What signs of a scam do you see in this message, and how should I verify it?" It is a useful second opinion, but always verify independently before acting. Never paste passwords or codes. For a wider view of the risks AI brings, read Challenges of Artificial Intelligence: 8 Real Problems, and for protecting your data when using AI tools see Data Privacy and AI Tools.

Related guides

Keep reading in this topic cluster:

• What Is Quantum AI? Separating Facts From Hype
• AI Transparency: How to Label AI-Generated Content
• Challenges of Artificial Intelligence: 8 Real Problems
• Data Privacy and AI Tools: What Happens to Your Data
• How to Use AI for Personal Finance (Safely)

Frequently Asked Questions

How can I tell if a voice call is AI-generated?

It can be very hard by sound alone. Hang up and call the person back on a known number, or ask a question only they would know. Treat urgent money requests as suspicious.

Are celebrity investment videos on social media real?

Videos of famous people promoting trading platforms or guaranteed returns are very often deepfakes. Legitimate public figures rarely endorse investment schemes in adverts.

Can deepfake detection tools spot every fake?

No. Detection tools help but make mistakes. Verifying through a separate trusted channel is more reliable than any detector.

What is a family code word?

A private word or phrase agreed in advance that family members use to confirm their identity in an emergency call or message.

What should I do first if I sent money to a scammer?

Contact your bank immediately through official channels, then change compromised passwords and report the scam to the police and the platform involved.