Still think you can spot AI? How to catch more convincing AI images, deepfakes and scams

AI images, videos and writing used to be relatively easy to spot. There were too many fingers, wonky backgrounds, weird junk text, and other visual artifacts that really quickly gave the game away to the AI. Deepfake videos often had delayed lip-syncing, and AI-generated writing felt repetitive and formulaic, like reading a clumsy LinkedIn post.

But AI has gotten better. A host of AI-generated images, videos and messages now look convincing enough to fool even the most careful observers and experts. Which means some of the older tricks for detecting what is AI and what isn’t work no longer as reliably as they once did.

So this isn’t about spotting signs and bugs. It’s about taking extra steps to confirm what you see and hear. We can no longer rely on zooming in to see six fingers in a weird-looking ad. We need to know what questions to ask and what tools to use when something doesn’t feel quite right.

Fact check on Facebook pages

Silhouette of smartphone with Facebook, Messenger, WhatsApp, Instagram, Oculus apps and blurred META logo on background

(Image credit: Shutterstock / mundissima)

AI-generated content is everywhere on Facebook, especially photos and videos designed to trigger a strong emotional response.

In May, a Full Fact investigation looked at pages sharing AI-generated stories about UK politicians and found that many were governed outside the UK, despite using British-sounding names. The posts contained heartwarming stories about politicians donating millions, rescuing dogs or helping sick children. None of that was true.

A useful tool in situations like this is Facebook’s page transparency feature. Found in a page’s profile, it can reveal where a page is controlled from, when it was created, whether it has previously changed its name and other clues about who is really behind it. If these details don’t match the image the page is trying to project, that’s often a good reason to be skeptical.

Finds scams on LinkedIn

Zoom Verified on LinkedIn Example

(Image credit: LinkedIn)

Job scams aren’t new, but artificial intelligence is making them harder to spot. Scammers can now generate realistic recruitment messages, professional-looking profiles and persuasive emails at scale.

Some recent campaigns have even mimicked LinkedIn messages and job alerts, using urgency and curiosity to push people to click on malicious links or share sensitive information.

Before you respond to an unexpected job offer, do some basic checks. Is the recruiter associated with a real company? Does their profile look legit? Is the company listed on Companies House? Taking a few minutes to confirm the option can save a lot of trouble later.

Verification of viral videos

Person using mobile phone trapped in a scroll hole surrounded by collage of social media obsessions.

(Image credit: Getty Images / We Are)

Deepfake videos become eerily realistic. At normal speed, an AI-generated video can now look completely authentic. But you can try increasing the playback speed, which is when subtle inconsistencies can become easier to spot. Look out for lip movements that don’t quite match the speech, unnatural blinks, strange movements, or facial movements that feel slightly out of sync with the sound you’re hearing.

It is also worth paying attention to facial expressions. Now, of course, no one responds perfectly all the time, but if a person’s expression consistently feels disconnected from what they’re saying, it might be worth investigating further.

Survey of AI influencers

Tilly Norwood, an artificial intelligence actor, sits in an office and looks into the camera

(Image credit: Xicoia)

Sometimes the answer is hiding in plain sight. You just have to know where to look. AI influencers, AI-assisted fashion shoots and AI-generated advertising campaigns are often revealed by brands and creators, but the information can be hard to find in a caption, hashtag, profile description or small print. For example, lifestyle publisher and brand SheerLuxe creates AI-generated content and influencers under the name Sheerluxe lab.

You can also try a reverse image search. Uploading a photo to Google Photos can sometimes show you where it first appeared online, whether it’s been modified, and whether the person in the photo actually exists elsewhere on the Internet. If an influencer only appears in AI-generated content and nowhere else, that’s a pretty big sign that they might be completely made up.