Seeing used to be believing. You could look at a photo or video and count on that being pretty good evidence that something really happened. If you heard a familiar voice on the other end of the phone line, you knew who was calling. A professional-looking website suggested somebody, at a minimum, put effort into being legitimate.

Even five years ago, none of this was foolproof. Photoshop has existed for decades and tools like WordPress made it easy to put together good-looking websites for a few hours of effort. But AI has made our old tools a lot less useful.

Scammers can produce convincing pictures and familiar voices in an instant. A website that looks like it belongs to a bank or government can be put together in an hour.

The old advice is “trust, but verify.” It’s still pretty good advice.

But how do you verify when the thing you’re trying to verify might not have even been made by a human being at all?

You probably can’t tell if something is AI.

If you’ve seen pictures with six fingers and an Escher staircase in a suburban home, then you probably figured out something was off. But I wouldn’t count on that. I can tell you for a fact, working online day in and day out, it is shockingly hard to tell AI text from human text and even AI images are getting pretty convincing.

This is not just me. A 2024 meta-analysis of 56 studies showed that more than 86,000 participants could only identify deepfakes about 55.5% of the time. That’s basically a coin flip. And AI tools have only gotten better since then.

You can’t base your security plan on noticing something weird. Take voice cloning for example. Our Scam Library describes a California mother, Deborah Del Mastro, who received a call from a “kidnapper” who claimed to have taken her daughter. She heard a recording of a voice sounding like her daughter in distress and wired $5,000 to Mexico. Her daughter was at work.

This could happen to anyone. Fear and urgency make it very hard to think straight, and it is trivially easy to spoof a phone number and fake a voice using common online tools.

In 2025, Americans reported losing $3.5 billion to impersonation scams, according to the Federal Trade Commission. That’s roughly three times what it was in 2020. Not all of that is due to AI, as the trends have been going this way for a long time, but it’s very hard to deny AI is part of the problem.

Official-looking sources can be faked.

This is particularly nasty. The FBI has warned about criminals impersonating its Internet Crime Complaint Center, or IC3, to target people who have already been scammed. (The unmitigated nerve of some people…)

Some use AI-generated videos of non-existent FBI personnel. Others contact victims over messaging apps and push them to fake websites to “update” their complaints. This is a ruse to collect more money and sensitive information from someone who is already vulnerable.

This works because it looks official, like a government agency, and there is a plausible reason for contact and what looks like supporting evidence. Everything looks like it fits together, except for the fact that the whole thing is a scam.

5 Ways To Verify Online Information

You don’t have to be a detective or a tech genius to spot scams. Here are five habits that you can build today to make yourself a little less vulnerable to scammers.

1. Use an independent channel. Imagine you get a text from an unknown number claiming to be your bank. They give you a phone number to call and a link to click. Don’t use either.

Open your banking app or call the number on the credit card. If a family member calls with an emergency, try reaching them through a number you already know. You might even try having a family codeword, but don’t rely on this as your only check.

2. Look at other sources. Researchers call this lateral reading and it’s a good habit even outside of cybersecurity. Don’t study a suspicious website more closely. Leave it and see what other sources say about the person, business, or claim. A study of 499 high school students found that teaching this technique significantly improved their ability to judge online credibility.

3. Trace claims to their origins. If you see a video, try to find out where it first appeared. See if a reputable news organization is reporting an event that you’re not sure about. Before you believe a photo, try to find out where it came from instead of relying on a screenshot with a dramatic caption. Remember that real footage can be used to tell false stories.

4. Don’t assume AI detectors will settle the question. For one, common detection tools are often wrong. There are newer systems like Content Credentials that can provide information on where a particular file came from, but they can’t tell you if the claim in the file is true.

5. When the stakes are high, research harder. You don’t need to verify every funny meme you see on Instagram. That would be a chore. But if someone wants your money, your password, or your private information, then scrutinize them much more carefully.

❝

If someone wants your money, your password, or your private information, then scrutinize them much more carefully.

Trust, But Verify

It’s tempting to look at how much AI slop is online and throw your hands up. That cynicism is easy to understand, but it won’t help you on Tuesday when you have things to do.

The better way forward is to get more demanding about evidence. Track down the sources of claims when you can and independently verify them whenever possible. You don’t need to be an expert in online security to make yourself safer. You already have tools you need. The only thing that needs to change is assuming that “looks legit” means “legit.”