Deepfakes, fake photos and fake reviews: check the source, not the pixels

A photograph used to be evidence. So did a video of someone speaking to camera. That has changed. The FBI warned in December 2024 that criminals now use AI to generate profile photos, videos of public figures and cloned voices, and that the technology “reduces the time and effort criminals must expend to deceive their targets”.

The word for this is a deepfake: in Ofcom’s definition (Ofcom is the UK communications regulator), audio-visual content “generated or manipulated using AI” that “misrepresents someone or something”.

Where you will actually meet one

The celebrity investment video. A familiar face on social media explains how they made a fortune from a trading scheme or cryptocurrency. Which? reported in August 2026 that these deepfake scams are common and have used AI images of Martin Lewis, Elon Musk, Bear Grylls and former Prime Minister Rishi Sunak, among others. MoneySavingExpert, the site Martin Lewis founded, says his face is used by scammers more than any other celebrity’s, and its rule is blunt: “if you see an ad with Martin that does not link back to MSE, it’s fake.”

The romance or friendship scam. The FBI says generated photos are used to build believable profiles for romance and investment fraud, and that deepfake video is used in fake video calls. The script is the old one the FTC (the US consumer protection agency) describes: someone who can never quite meet in person and eventually needs money.

The “official” on a video call. In a warning issued in September 2026, the FBI said scammers posing as police officers or government officials now use AI to appear as those officials on video calls, “adding the appearance of legitimacy to the scam”. It counted nearly 61,000 complaints about law enforcement or government impersonation scams between January 2025 and July 2026, with losses of more than $1.6 billion; those figures cover these impersonation scams as a whole. The FBI’s advice is that law enforcement and government authorities will never contact you by telephone or text message to demand payment, and will never ask to be paid by prepaid card, cryptocurrency or courier. To check, find the agency’s official, publicly listed contact details yourself and use those, not the ones the caller gives you. The warning is American, but the check works anywhere.

The job interview with nobody there. In March 2026 AARP, the US membership organisation for people aged 50 and over, described a 65-year-old communications executive who applied for a well-paid remote job and was interviewed by text on WhatsApp by what turned out to be a chatbot. It responded to his answers and asked him to say more. A 10-page job description followed, then a 30-minute phone interview with a synthesised voice, and a job offer within three days. The speed made him suspicious, and he broke off before any harm was done. AARP’s advice is to slow down, check independently and guard your personal details. A real employer can be reached through the contact details on its own website. Our guide to job hunting covers the rest.

The fake review. Plausible five-star write-ups, which AI can now generate to order.

The breaking-news image. The Associated Press notes that AI content makes it “increasingly difficult to separate fact from fiction, particularly when it comes to breaking news”.

Why looking for glitches is not enough

For a while the advice was to look for six fingers, odd teeth or strange blinking. The FBI’s 2024 notice still lists “distorted hands or feet, unrealistic teeth or eyes”. But the Associated Press, in an April 2026 guide, says that as the technology evolves “these clues aren’t as common as they once were”. Worth a glance, and nothing more.

What a forger cannot easily fake is where the thing came from. So check the source, not the pixels.

The method

There is a named version of this, called SIFT, developed by the digital literacy expert Mike Caulfield and taught by the University of Chicago Library, among others: Stop, Investigate the source, Find better coverage, Trace claims to their original context. In everyday terms:

1. Stop. Before you share, reply, click or pay. The Chicago guide points out that headlines are often written to get clicks by provoking “a strong emotional response”.
2. Ask who posted it. Then look on the person’s or organisation’s own official website or account. A real announcement will be there.
3. Look for better coverage. Search for the claim on a news site you already trust. If a remarkable event appears nowhere else, treat it as unconfirmed.
4. Trace the image. A reverse image search means giving a search engine the picture instead of words, so it shows you where else that picture appears. Google’s help page says you can upload an image, or right-click one in Chrome and choose “Search with Google Lens”, to find websites with the same or a similar image. For a video, the AP’s tip is to take a screenshot first. The FTC’s romance-scam advice is to do this with the person’s profile picture.
5. Be most suspicious when it wants something. Money, urgency or a click.
6. Ask someone else. A second pair of eyes costs nothing, and scams rely on you deciding alone.

For any investment you see advertised, the UK’s Financial Conduct Authority (FCA), which warns that trading scams “often use fake celebrity endorsements”, says to use its Firm Checker and Warning List. There is more in Pension and investment scams after 50. This is general information, not financial advice.

Fake reviews: the tells and the law

The FTC is candid: “Don’t assume that, just by looking, you can spot the difference between a real review and a fake one.” Its practical checks are about pattern, not prose:

• A burst of reviews over a short period.
• Reviewers whose account seems to exist for one review of one product.
• A label or badge showing the reviewer received an incentive.
• Fakes run both ways: some give less than top marks to seem believable, and some are negative ones posted to harm a competitor.

It also says to look at a variety of sources, including well-known websites with trustworthy and impartial expert reviews.

The law has caught up. In the UK, the Digital Markets, Competition and Consumers Act banned fake reviews from 6 April 2025, Which? reported, and sites must take reasonable steps to check that reviews are genuine. The Competition and Markets Authority can fine a business up to 10% of global turnover, and in March 2026 it announced investigations into five businesses. In the US, the FTC announced a final rule in August 2024 banning fake reviews and testimonials, explicitly including AI-generated ones, with civil penalties for violators.

When you find one

Report it on the platform where you saw it. Then, for the scam versions:

UK: Report Fraud, the City of London Police service that replaced Action Fraud in December 2025, at reportfraud.police.uk or on 0300 123 2040; in Scotland, Police Scotland on 101. Report suspect investment firms to the FCA on 0800 111 6768. If money has left your account, ring your bank as well.

US: ReportFraud.ftc.gov, and the FBI’s Internet Crime Complaint Center at ic3.gov.

This week

Keep some proportion. Most of what you see online is real: real photographs, real people, real opinions about real kettles. The aim is one habit, not suspicion of everything, applied at the moments that matter: before you share, buy or pay, ask where it came from.

For the wider picture of using AI well, start with AI after 50: the complete guide.

Updated 5 October 2026: added the FBI’s September 2026 warning about scammers using AI to pose as police or government officials on video calls.

Updated 7 October 2026: added AARP’s March 2026 account of an AI chatbot running a fake job interview.