How to Spot AI Scams, Deepfakes, and Fake Content

Spot visual "glitches" in photos

AI image generators are computer programs that create pictures based on text descriptions, much like a highly skilled digital painter who has studied millions of real photos. While they're incredibly talented, older versions of these programs often struggle with fine details, and examining photos closely can sometimes reveal inconsistencies.

When you see a photo that feels "off," your first action should be to zoom in and closely examine the specific details. You'll usually see the image take up more of your screen as you zoom. Once magnified, check areas like hands, feet, accessories, clothing, and backgrounds for subtle errors. If your device doesn't have a simple pinch-to-zoom feature, look for a magnifying glass icon or a dedicated zoom option in your photo viewer. You'll know it worked when you can clearly see distorted features, strange shadows, or illogical elements in the image.

Note: Modern AI generators have improved dramatically, so these artifacts are becoming less common. When in doubt, verify the image through a reverse image search or other means.

Identify "deepfake" video clues

A deepfake is a video where a person's face or body has been digitally altered or even completely swapped using smart computer systems. Think of it like a sophisticated digital mask that moves in real time, trying to perfectly mimic a real person. While detection is getting harder as the technology improves, there are still some things to watch for.

Start by playing the suspicious video, then pause it frequently to get a good look at specific frames. As you play the video, you'll see the frames move, and then when you pause, a still image will appear, allowing you to scrutinise the facial areas without the distraction of constant motion. If your video player doesn't have an obvious pause button, look for a large play/pause icon, often in the center or bottom of the screen. You'll know it worked when you spot any inconsistencies in how the face blends with the hair, unusual lighting, or areas that look slightly blurred or distorted.

Note: High-quality deepfakes are increasingly difficult to spot with the naked eye. If a video requests money or sensitive information, always verify through a separate, trusted method.

Listen for voice cloning issues

Voice cloning is technology that takes a short recording of a real person's voice and then uses a computer to make them say entirely new things. Imagine it like a high-tech text-to-speech tool that can perfectly imitate your friend, child, or colleague. Modern voice cloning can be very convincing, so the best protection is not relying on voice alone.

When you receive a suspicious call or audio message, listen intently to the speaker's voice, paying close attention to their tone and speech patterns. You'll hear the words being spoken and can analyze how natural or unnatural they sound. If you are on a call, try to get to a quiet place or use headphones if available to reduce background noise. If anything feels slightly off — an unusual tone, unexpected phrasing, or a request that feels out of character — take that as a signal to verify through other means.

Verify suspicious messages using a second channel

Deceptive messages often leverage AI to write highly convincing, personalized emails or text messages, making them a potent form of automated scamming. The golden rule of online safety is simple: never use the contact details provided within a suspicious message to verify its authenticity, as these will likely lead you directly to the scammer.

Your action is to deliberately step away from the original message and initiate contact through a different, independently verified method. You'll see the original message on your screen (email, SMS, social media), which might contain urgent requests or "updated" details. By using an independent contact, you avoid giving the scammer direct access. If the request was via email, for instance, don't reply to that email. Instead, find a trusted phone number from a previous bill or the company's official website. You'll know it worked when the person on the other end confirms (or denies) the request, using a contact method you initiated and trust, separate from the potentially fake one.

Look for generic or overly perfect text

AI-generated text can often be identified not by obvious errors, but by its perfection – or lack of human nuance. These AI-written texts might be overly formal, incredibly verbose, repetitive in their phrasing, or simply too generic to sound like genuine human communication. They lack the quirks, emotional inflections, and subtle imperfections that make human writing unique.

Your action is to read the message or article carefully, paying close attention to its style, flow, and word choice. As you read, you'll be evaluating the tone and identifying any unusual patterns. This applies whether you're looking at an email, a social media post, an online article, or even a review. If the text reads like it was written by a robot trying to sound human, it's a good clue. You'll know it worked when the writing feels unnaturally smooth, free of human quirks or slang, or uses overly formal language for a casual situation.