Every day, millions of Americans receive suspicious emails and text messages designed to steal their money or personal information. What’s changed recently is that artificial intelligence is making those scams harder to spot — and easier to send by the millions.
Brian Watson, community outreach specialist for the nonprofit ROSE: Resources Outreach to Safeguard the Elderly, works regularly with communities across southern Arizona to help people recognize and avoid scams. In a conversation on The Show from KJZZ, Watson explained how AI is reshaping the scam landscape — and what ordinary people can do to stay safe.
The Biggest Change: Phishing Emails Look Legitimate Now
For years, a tell-tale sign of a phishing email was poor grammar, awkward phrasing, or spelling errors — the result of scammers writing in a language that wasn’t their own. That red flag is largely gone. Watson, who also teaches at the University of Arizona, demonstrated this to his students by running a poorly written 2016 phishing email through ChatGPT.
“It was absolutely amazing how much just a few seconds running it through AI cleaned it up to the point where it was actually believable. And this was a poorly written phishing email from 2016,”
Watson said.
The implications are significant. Scammers no longer need strong English-language skills to craft convincing messages. AI tools polish their writing instantly, making fraudulent emails nearly indistinguishable from legitimate ones.
One Scammer, Millions of Messages
Beyond polished language, AI has dramatically increased the scale at which scammers can operate. Watson explained that automated bots now allow a single bad actor to blast out massive volumes of fraudulent messages with minimal effort.
“You can have one person utilizing AI, and that one scammer can send out thousands, if not millions, of phishing emails or wrong number texts. And they just send them out like an army of bad actors,”
Watson said.
“So, instead of them having to type up and send a bunch of emails one by one or copying and pasting them, these bots that they use automate it.”
This efficiency factor is what Watson calls the most important development. The same AI tools that help everyday users work faster and smarter are being exploited by criminals for the same purpose. As Watson put it,
“AI makes scammers a lot more efficient.”
Voice Cloning and Deepfakes: A Growing But Labor-Intensive Threat
AI’s role in visual and audio deception is also expanding, though Watson notes it remains more targeted in practice. He has seen demonstrations of voice cloning technology — where just a sentence or two of someone’s recorded voice can be used to generate entirely new speech — at presentations with local sheriff’s departments in southern Arizona.
While the technology is striking, Watson points out it requires significant effort to deploy effectively. In schemes like the “grandparent scam,” where fraudsters impersonate a grandchild in distress, voice cloning only works if the target actually answers the phone. A missed call means wasted work.
Deepfake video, however, is finding a practical use in romance scams. Victims who insist on a video call — often as a way to verify the identity of someone they’ve met online — can now be fooled by AI-generated video.
“The scammers can actually do the deepfake videos and make it look more legitimate,”
Watson explained, noting that most romance scammers still prefer to keep contact limited to text or messaging apps whenever possible.
Most Scams Still Don’t Rely Heavily on AI
Despite widespread public concern about AI-powered fraud, Watson offers a more measured perspective. At every presentation he gives, audience members ask about AI — but his answer is consistent: most scams don’t depend on it heavily.
“Most scams actually don’t rely heavily on AI. I mean, a lot of scams, like phone scams, are old school technology. They’re just executed so well,”
he said.
This mirrors a broader pattern Watson describes as a “cat-and-mouse game” between criminals and law enforcement. Every new technology — including cryptocurrency, which Watson cites for its anonymity benefits to bad actors — gets picked up by both sides. AI and data centers are simply the latest chapter in that ongoing dynamic.
What Regulations Can and Can’t Do
Conversations about government regulation of AI are growing louder, including discussions about whether new rules could help curb AI-assisted scams. Watson is realistic about the limits of that approach.
“Whatever rules you put in place and laws, the scammers are not going to obey because they’re scammers by nature, especially if they’re overseas in another country,”
he said. Regulations, he noted, tend to shape the behavior of legitimate companies — not criminals.
That means the responsibility ultimately falls back on individuals.
“It’s always going to come back to consumer awareness and protecting ourselves and just being very cautious,”
Watson said.
Practical Steps to Protect Yourself
With the old spelling-and-grammar filter no longer reliable, Watson recommends building a new mindset around any unsolicited digital communication. His advice centers on healthy skepticism rather than fear.
For any email from an unknown sender, Watson suggests asking a few key questions: Could this be a scam? Where did it actually come from? Is it asking for personal or financial information? Does it create a false sense of urgency?
He also offers one surprisingly simple tip: don’t act on suspicious messages from your phone.
“Your smartphone is so, that screen is so small. Wait till you get home, look on a big screen, take your time, and a lot of times you’ll figure out that it is a phishing email,”
Watson said.
Slowing down, Watson argues, remains one of the most powerful defenses available — even as the technology used against consumers keeps accelerating. For more on how AI is intersecting with everyday life, including topics like whether AI should hold a medical license and the politics of AI in Washington, reporting continues from newsrooms tracking these developments closely. Locally, questions about the growth of data centers in Arizona and Arizona’s semiconductor investments are part of the same expanding conversation about how AI infrastructure is reshaping communities.
Watson’s broader message: the tools scammers use may be getting smarter, but so can the people they’re targeting — starting with slowing down and questioning what lands in your inbox.
