For years, security training taught people to spot phishing by its mistakes: clumsy grammar, odd spelling, a generic "Dear Customer." That advice is now out of date. With AI writing tools, attackers can produce flawless, natural-sounding emails in any language, in seconds, tailored to the person reading them.
AI also makes personalization cheap. An attacker can pull your company's website, staff profiles and recent news, and write a message that mentions a real project, a real vendor and a real coworker's name. It reads exactly like a normal day at work.
So the question to train people on is no longer "Does this email look sloppy?" It's "What is this email asking me to do, and how can I check it's real?"
The new rule of thumb: judge a message by its request, not its writing. Polished writing proves nothing. A request involving money, passwords, sign-ins or sensitive data always gets verified, no matter how real the message looks.
An example, annotated
Here's the kind of message we mean. It has no spelling mistakes, a plausible story and a believable sender. The company and names are invented for this example.
Hi Jordan,
Thanks again for getting the Q3 order through so quickly. Following up on the thread below: our bank is moving us to a new account as part of a merger, so please use the updated details on the attached form for invoice 44817 and any payments going forward. 3
We need this processed by end of day today to avoid a late fee on our side. I'm in back-to-back meetings, so email is best for now. 4
View updated remittance form 5
Best,
Dana Whitfield
Accounts Receivable, Northside Supply
- 1Look-alike domain. "suppIy" uses a capital I instead of a lowercase L. On many screens, they're indistinguishable. Attackers also use extra hyphens, swapped letters, or different endings like .co instead of .com.
- 2"RE:" with a real invoice number. Replying into an existing conversation (sometimes from a vendor's genuinely hacked mailbox) makes a message feel trusted. A "RE:" proves nothing.
- 3A change to payment details. This is the single biggest red flag in business email. Legitimate bank changes are rare and should always be confirmed by phone.
- 4Urgency plus a reason you can't call. A deadline, a penalty, and "I'm in meetings, email only" are designed to stop you from verifying.
- 5A button leading to a "form". These often lead to a fake sign-in page that captures your Microsoft 365 password and MFA approval.
The new red flags
Ignore the grammar. Look for these instead:
01A request involving money, credentials or data
Changing bank details, paying an urgent invoice, buying gift cards, updating payroll direct deposit, sharing a password or MFA code, or sending a file of customer or employee data. The more sensitive the request, the more verification it needs.
02Pressure, secrecy or flattery
"Needs to happen today." "Keep this between us for now." "You're the only one I trust with this." Real requests can be urgent, but genuine colleagues don't object when you double-check.
03A sender address that's almost right
Check the actual email address, not just the display name, which anyone can set to anything. On a phone, tap the name to reveal the address. Also check whether replies would go to a different address than the sender's.
04A sign-in request you didn't start
"Your password expires today." "You have a secure document waiting." "Re-verify your account." Any link that leads to a sign-in page is suspect. Go to the service directly by typing its address or using a bookmark instead.
05Unexpected attachments and shared files
Shared documents, e-signature requests, voicemail files and "scanned invoices" you weren't expecting are common lures, even when they come from a familiar-looking file-sharing service.
06A push to switch channels
"Text me on my personal number." "Let's move this to WhatsApp." Moving the conversation off company systems removes the protections and records that would catch the scam.
07Something that's slightly out of character
The CEO who never emails you directly suddenly does. A vendor who always calls is now email-only. The tone is a little more formal, or more casual, than usual. Trust that "huh, that's odd" feeling. It's often the only clue.
Beyond email: voice, video, text and QR codes
AI isn't just writing emails. The same red flags apply across every channel:
- Voice cloning: a short clip of someone's voice, from a video, a podcast or a voicemail greeting, can be enough to imitate them on a phone call. A call that sounds like the owner asking for an urgent wire transfer still gets verified.
- Video calls: live deepfake video is getting easier. Be wary of urgent financial requests on a video call, especially if the camera quality is poor or the person avoids questions.
- Text messages (smishing): fake delivery notices, toll payments, bank alerts and "this is your boss, are you free?" texts.
- QR codes (quishing): QR codes in emails or on printed notices that lead to fake sign-in pages. Scanning on a personal phone also sidesteps your company's email and web protections.
- Fake IT support: calls or chat messages from "the help desk" asking you to approve an MFA prompt, install a remote-access tool or read out a code. Real IT will never ask for your password or an MFA code.
The Pause–Check–Verify routine
Give your team one simple habit to use whenever a message asks them to do something sensitive:
| Step | What to do |
|---|---|
| 1. Pause | Don't act right away, even if it's urgent. Urgency is the attacker's main tool. Five minutes won't hurt a real request. |
| 2. Check | Look at the request, the real sender address, and where any link actually goes (hover over it on a computer, or press and hold on a phone). |
| 3. Verify | Contact the person through a channel you choose: a phone number you already have on file, the company directory, or walking over to their desk. Never use the phone number, link or reply address in the message itself. |
And if it still seems off, report it. Use your email's "Report phishing" button or forward it to IT, then delete it. Reporting helps protect everyone else who received the same message.
What to do if someone clicked
People will occasionally get fooled. That's expected. What matters is reporting it fast, because the first few minutes make a big difference. Make sure everyone knows to:
- Tell IT immediately, even if they're embarrassed or not sure anything happened.
- Say what they did: clicked a link, entered a password, approved an MFA prompt, opened an attachment, or sent money or data.
- Not try to fix it alone by deleting emails or running tools. IT may need the evidence.
- Contact the bank right away if a payment was sent. Acting quickly gives the best chance of stopping or recovering a transfer.
Never punish someone for reporting. If people fear getting in trouble, they'll stay quiet, and a five-minute fix becomes a week-long incident. Thank people for reporting, especially when they made a mistake.
What leaders should put in place
Training helps, but the strongest protections are processes that don't depend on someone spotting a fake:
- A payment-verification rule: every new payee and every change to bank details is confirmed by phone using a number already on file, and large or unusual payments need a second approver.
- A verification code word for urgent requests made by phone or video, known only to the people who'd need it.
- Phishing-resistant MFA such as passkeys, so a stolen password and a fake sign-in page aren't enough to break in.
- Strong email filtering and domain protection (SPF, DKIM and DMARC), plus "External sender" banners.
- Short, regular training with realistic phishing simulations, instead of one long session a year.
- A one-click way to report suspicious messages, and a quick, friendly response when people do.
Quick reference for your team
Print this section and post it near desks, or share it in your team chat.
Phishing red flags
When in doubt: Pause. Check. Verify using a number you already have. Report it.
The bottom line
AI has taken away the easy giveaways, but it hasn't changed what attackers want: money, passwords and data. Teach your team to focus on the request instead of the writing, to verify through a channel they choose, and to report without fear, and you'll stop most of these attacks no matter how polished they look.
Want help preparing your team? We offer security awareness training and realistic phishing simulations, plus the email and sign-in protections that catch what people miss.