AI Phishing Emails: What Still Gives Them Away When the Typos Are Gone
AI phishing emails are scam messages written with the help of an artificial intelligence tool, and they no longer carry the clumsy wording that used to give fraud away. For a small business, that changes one thing above all: the advice to look for spelling mistakes now protects almost nobody. What still gives these emails away is what they ask the reader to do, and that is where your team’s attention should go.
This article explains what changed, walks through a real case reported in October 2026, and ends with the questions to put to your team and to your IT provider.
The Spelling Mistake Test Has Stopped Working
For years, security training taught employees to spot fraud by its form: a strange greeting or a broken sentence, or a sender who wrote like a machine translation. That advice worked because writing a convincing email in a second language took time, so criminals reused the same templates and the same awkward phrases.
AI tools removed that constraint. Today, anyone can produce a polished message in a few seconds, in any language, matched to the role of the person receiving it. The same tools can copy a writing style and invent a whole earlier conversation to make a request look routine.
In short, the email your receptionist reads on Monday morning can be perfectly written and still be a scam.
One Person, More Than 30 Different Emails
A case presented in October 2026 by an email security company, and written up in The Hacker News, shows what this looks like in practice.
One person contacted more than 30 agents at the same real estate company. Each message came from a different, real email account. Each one asked about a different property, in different words. The questions were ordinary: whether a house was still open for viewing, what the asking price was, whether a visit was possible that week.
In one thread, an enquiry sent at 2:58 in the morning turned into an invitation to a video meeting 13 minutes later, with a branded template and a note saying the meeting needed the agent’s immediate attention. Every conversation, whatever its opening question, ended with the same request: join this video call.
For the agents, nothing looked unusual: a buyer asked about a listing and wanted to talk, which is exactly what an agent hopes for.
Why the Email Filters Let It Through
Most email protection works by comparison. It checks whether a sender is known, whether the wording matches messages already flagged elsewhere, and whether a link points to an address with a bad reputation. That approach still works well against fraud built from reused templates and known bad addresses.
This campaign slipped past all three checks. Every sender was writing for the first time, so there was no history to compare against. Every message was different, so there was no repeated text to match. And the meeting link began with a genuine Google address, so the reputation check came back clean, even though the link then sent the agent to an unrelated site.
That is why the attack worked: each email, read on its own, looked like ordinary business. The pattern only appears when somebody looks at all 30 messages together, and no single agent ever saw more than one.
AI Phishing Emails Are Becoming More Common
The trend shows up in the large breach studies as well. In its 2025 Data Breach Investigations Report, Verizon found, using data from one of its partners, that the share of AI-written text in malicious emails had doubled over the previous 2 years.
Read that figure honestly. It comes from one partner’s view of malicious emails sent to companies of every size, and it measures how often AI helped write them. It does not measure how many of those emails succeeded, and it does not say your company receives the same mix. What does transfer to a 40-person business is the direction: the writing in scam emails keeps getting better, and the tools that make it better are cheap and available to anyone.
What AI Cannot Hide: The Request
An attacker can polish every sentence of an email. What they cannot change is the reason they sent it. Every scam has to end with the reader doing something, because that action is the only way the attacker gets paid.
The request usually takes one of a few forms:
- Click a link or open a document
- Sign in on a page that looks familiar
- Join a video call or a meeting
- Change bank details or send a payment
- Keep the exchange private, or deal with one person only
This is the part of AI phishing emails that deserves your team’s attention. A polite question about a listing is harmless. A polite question that turns, within minutes, into a meeting link from someone you have never met is a different message, however well it is written.
The security company behind the real estate case describes its own response the same way: read each message for what it is trying to achieve, and whether that goal makes sense for this sender and this conversation. You do not need their product to apply the principle. Your team can apply it with nothing more than a habit.
Businesses That Answer Strangers for a Living Are the Easiest Targets
The pattern behind AI phishing emails like these works on any company whose daily work depends on hearing from people it has never met. A clinic taking new patients, an accounting firm signing new clients, an insurance agency quoting policies or a contractor answering requests for bids all fit that description.
For these businesses, an email from a stranger is simply part of the job, which is exactly why the request at the end of the conversation matters more than who the sender is or how well they write. According to The Hacker News write-up, a variant aimed at insurance agents built trust over several exchanges, with quote requests and coverage questions, before sending a fake conference invitation.
Three Questions Your Team Can Ask About Any Email
None of this requires new software. It requires a short habit that anyone on your team can apply in a few seconds, especially the people who answer the general inbox, the front desk and the sales line.
- What is this message asking me to do? Name the action out loud: click, sign in, join a call, pay, change details.
- Would this person normally ask for that, and this quickly? A first contact that jumps to a meeting link or a payment within minutes deserves a pause.
- Can I check it through a channel I already trust? Call back on a number you already have, look the person up through your own records, or ask a colleague before acting.
If the answer to the third question is no, the request waits. A real buyer, client or supplier will still be there an hour later.
Questions to Ask Your IT Provider
Your provider handles the technical side. These questions tell you whether that side has kept pace with AI phishing emails.
- What does our email protection do with a polite message from a brand-new sender that ends with a meeting link? A good answer describes how links are checked when someone clicks, and not only when the email arrives.
- Would you notice if the same request reached several of our employees from different senders? This is the pattern that the real estate campaign relied on.
- If someone on our team signs in on a fake page, how quickly would you know, and what would you do first? You want to hear about cancelling open sign-ins and checking the mailbox for hidden forwarding rules.
- How do our employees report a suspicious email, and how fast do you respond? A report button that nobody answers teaches people to stop reporting.
If Someone Already Clicked
Ask employees to say so immediately, and make it clear they will not be blamed. A click reported within minutes is usually cheap to fix, while one hidden for a week can cost a great deal.
Then call your IT provider. Ask them to cancel the employee’s open sign-ins, reset the password, check the mailbox for rules that forward or hide messages, and look for the same email in other inboxes. If money moved, call your bank right away and file a complaint with the FBI at ic3.gov.
The Habit That Outlasts Better Writing
AI phishing emails will keep getting better written, and no training can teach people to spot a flaw that is no longer there. What stays visible is the request. A team that asks what an email wants before it acts, and checks that request through a channel it already trusts, is far better prepared for the next generation of scams as well as this one.
