82% of phishing emails are now written by AI

Today's Tip: Spotting Scam Emails Before You Click

Everyone checks scam emails backwards. They squint at the message, look for typos, and hope for the best… meanwhile AI can tell you in 30 seconds whether you're about to get robbed.

Let me explain…

My friend Marvin called me last Tuesday morning sounding rattled. He's 68, retired postal worker, sharp as a tack — not the kind of person you'd peg as scam bait. But there it was in his inbox… a message that looked exactly like a FedEx delivery notice. A package "couldn't be delivered." Click here to reschedule. Pay a $3.45 redelivery fee.

The thing is, Marvin WAS expecting a package. A new pair of orthopedic shoes from his daughter. The timing fit. The logo looked right. And $3.45 felt small enough to not bother arguing about.

That's how they get you. Not with the obvious "Nigerian prince" nonsense from twenty years ago… but with a small, plausible, perfectly-timed request for almost nothing.

Here's what's changed. Scammers now use AI to write these emails. The grammar is perfect. The tone matches a real company. The little tells we used to look for — misspellings, weird capitalization, broken English — those warning signs have mostly vanished. According to one recent industry report, 82% of phishing emails are now AI-generated, and elder fraud losses jumped 43% in a single year to nearly $5 billion.

But here's the good news. The same AI that's writing those emails can also catch them. And one tool in particular — because it can read the live web in real time — is uniquely good at this job. I'll get to which one in a minute.

First, here's exactly what I told Marvin to do… and what saved him from clicking that link.

★ Copy-Paste Prompt

Works in Claude, Gemini & Grok

I received the email below and I'm worried it might be a scam. Please act as a fraud analyst and review it for me.

Check the following and explain each one in plain English a non-technical person can understand:

1. The sender's email address — does the domain match the real company it claims to be from?

2. Any links in the email — what is the actual destination URL, and is it the legitimate site for that company?

3. The language and urgency — does it use pressure tactics, threats, or requests for immediate payment?

4. The request itself — is the company asking for something a real company would never ask for by email (passwords, full Social Security numbers, gift cards, wire transfers)?

5. Any other red flags you spot.

At the end, give me a simple verdict: SAFE, SUSPICIOUS, or ALMOST CERTAINLY A SCAM — and tell me exactly what to do next. If you can search the live web, please look up the real company's official contact information so I can verify on my own.

Here is the email:

[PASTE THE ENTIRE EMAIL HERE, INCLUDING THE SENDER ADDRESS AND ANY LINKS]

Why this works: You're giving the AI a specific job (fraud analyst), a numbered checklist so it can't skip anything, and a clear output format (verdict plus next steps). The numbered list is what makes the difference. Without it, the AI gives you vague "be careful" advice. With it, you get a real inspection of every part of the email.

When Marvin ran his FedEx email through this prompt, the answer came back in about 20 seconds. The sender's address wasn't fedex.com — it was something like "fedex-notify-secure.shipping-track-portal.com." The link in the email pointed to a fake page in a country that has nothing to do with FedEx. And the real FedEx, the AI explained, never charges redelivery fees by email. They leave a slip on your door.

Verdict: ALMOST CERTAINLY A SCAM. Action: delete the email, do not click anything, and if you're expecting a package, go directly to fedex.com and type the tracking number yourself.

Total time from "is this real?" to "I know exactly what this is": under a minute.

★ Today's "Aha!" Moment

You don't need to become an expert at spotting scams. You need a second set of eyes that's already an expert… and now you have one in your pocket, free, available at 2 a.m. when a suspicious text shows up.

I know what you're thinking. "Isn't it dangerous to paste a suspicious email into an AI? What if there's something bad in it?"

It's a fair question, and I respect it. But here's the answer. Pasting the TEXT of an email into a chat window is not the same as clicking the link. The danger in a phishing email is the link — clicking it takes you to a fake site that steals your password or installs something nasty. Copy-pasting the text into a chat box does none of that. You're just showing the AI words. It can't be tricked into clicking for you.

The one thing to be careful about: don't paste personal information INTO the email along with it. If the scammer's email already has your name in it, that's fine. But don't add "by the way, my account number is 12345" to your message. Keep the conversation focused on the scam, not on your details.

★ One Thing to Watch

AI is excellent at spotting obvious scams and verifying URLs. But it's not perfect, and it can occasionally be wrong on a borderline case. Treat its verdict as a strong second opinion, not as the final word. If the AI says SUSPICIOUS or you're still not sure, the safe move is the same as it has always been: do not click any link in the email. Open a new browser tab, type the company's real website by hand, and log in there. If something needs your attention, you'll see it inside your account.

Start using this today. Next time something shows up in your inbox or as a text that gives you that little pause — the half-second of "wait, is this real?" — don't squint at it and guess. Don't ask your kids and feel embarrassed. Open up an AI chat, paste it in, run the prompt above.

It costs nothing. It takes one minute. And it could save you thousands.

★ ★ ★

Tool Spotlight ★

Grok

Site: grok.com

Made by: xAI

Cost: Free tier available

Grok is the AI tool I'd grab first for today's job. The reason is simple: Grok pulls live information from the web as it answers you. So when you paste in a suspicious email, Grok can actually go check whether "amaz0n-customer-service-billing.net" is a real Amazon site (it isn't), look up Amazon's actual customer service phone number for you, and compare the email's claims against what's happening right now. The other AI tools can guess… Grok can verify.

Sign up takes about two minutes at grok.com using an email address. The free tier gives you enough chats per day to handle any normal week's worth of suspicious emails. Ask it anything that depends on current information — "is this charity that just emailed me a real one?" or "what's the current scam tactic involving Medicare cards?" — and you'll see why its live-web ability stands out from the rest.

Best for: verifying URLs and sender addresses, current events research, product comparisons, news fact-checking, looking up real company contact info, anything that needs up-to-the-minute information

★ ★ ★

★ Quick Wins

The "Translate This Doctor-Speak" trick — Next time you get a lab result or test summary that reads like a foreign language, paste the whole thing into Claude or Gemini and type Please rewrite this medical report in plain English a person with no medical background can understand. After the rewrite, list any results that are outside the normal range and explain what each one might mean. You'll get a version you can actually read. Then bring your questions to your next appointment instead of nodding politely.

The "What's a fair price?" check — Before you say yes to any quote — plumber, roof repair, car mechanic — open Grok and type I just got a quote of $[amount] for [the specific job, including parts and labor described as the contractor listed them] in [your zip code]. Is that a fair price for this area in 2026? What questions should I ask before agreeing? Because Grok reads the live web, it pulls current pricing from contractor forums and recent reviews. You'll know in a minute whether you're being quoted fairly or fleeced.

The "Reverse phone number" lookup — Got a missed call from a number you don't recognize, and you're nervous about calling back? Open Grok and type I got a missed call from [phone number]. Can you search the web and tell me who this number belongs to and whether it has been reported as a scam? Please cite where you found the information. Grok's live search will check scam-tracking sites and recent reports. Saves you from calling back a robocaller — or worse.

Skill Builder: Knowing When NOT to Use AI

Here's something most AI articles won't tell you. The smartest thing you can learn about AI isn't a clever prompt or a fancy trick. It's knowing when to put the AI down and pick up the phone instead.

AI is fantastic for some things. Drafting a letter, summarizing a long document, comparing two products, checking a suspicious email. It is genuinely not the right tool for others. And mixing those up is how people get into trouble.

When should you NOT use AI? Three situations come up again and again. First, anything that requires a licensed professional's judgment on YOUR specific case — diagnosing a symptom, drafting a will you'll actually sign, deciding how to invest a 401(k). AI can help you prepare questions and understand the vocabulary. It cannot replace your doctor, lawyer, or financial advisor. They have a license, malpractice insurance, and a duty to you. The AI has none of those.

Second, anything emotional or relational. If you're working through grief, a marriage problem, or a fight with your adult child, the AI will give you a polite-sounding answer that may not be wrong, but won't be right either. It doesn't know you. It doesn't know the people involved. A friend, a counselor, or a clergy member is the better call.

Third, anything where you need to verify a specific number, date, or fact and being wrong has real consequences. AI tools can still make up facts that sound convincing. If you're using a number in a tax return, a legal document, or anywhere it will be checked against the truth, look it up at the official source. Use the AI to help you find that source… don't use it as the source itself.

Here's a prompt you can use any time you're not sure whether AI is the right tool for what you're trying to do.

★ The "Is This Your Job?" Prompt

Works in Claude, Gemini & Grok

I'm thinking about using you to help with the following situation: [DESCRIBE WHAT YOU WANT TO DO]

Before we start, please answer three questions honestly:

1. Are you actually the right tool for this, or would a licensed professional (doctor, lawyer, accountant, financial advisor) serve me better?

2. What parts of this can you help with safely, and what parts should I take to a human expert?

3. If you proceed to help, what specific limitations should I keep in mind about your answer?

Please be candid. I would rather hear "you should talk to a professional" than get an answer that sounds good but might be wrong on something important.

Why this works: You're inviting the AI to flag its own limitations BEFORE it gives you advice. Most people skip this step and treat every AI answer as equally trustworthy. This prompt forces a moment of honesty up front. You'll be surprised how often a good AI will say "honestly, this one needs a professional" — and that's exactly the answer you needed.

Think of AI the same way you think of a sharp neighbor who happens to know a lot. A great resource for "help me think this through" and "show me how this works." A terrible resource for "tell me exactly what to do with my health, my money, or my family." The wisdom is in knowing which question you're actually asking.

Used this way, AI becomes a tool that makes you smarter — not a replacement for the people you trust.

82% of phishing emails are now written by AI

Read More from the AI Advantage