Short answer
AI-written phishing emails are harder to catch because they no longer contain the typos and awkward phrasing people were taught to look for. They can read like a real contractor, client or consultant wrote them. They may also mention real projects found on a firm’s website. The better defense is checking the sender address, avoiding sign-in links in unexpected emails and verifying any money or bank change by phone. Coverage for resulting losses varies by policy, so review the actual wording.
Why AI-written phishing emails look more convincing
For years the standard advice was to look for bad grammar. That tip is much less useful now. Writing tools can produce clean, natural messages in any tone, and they can adjust the wording for a specific person or company.
Four things have changed:
- Grammar. Spelling and punctuation errors are far less common.
- Tone. Messages can match the way a project team normally writes.
- Personalization. Details such as project names and consultants can come from public websites and social media.
- Timing. An email can arrive right before a real deadline or meeting.
What an AI-written phishing email can look like
Here is an invented example. It uses no real names or projects.
Subject: Updated drawings for the 4th floor revisions
Hi Jordan, please review the attached updates before Thursday’s OAC meeting. The files are on our shared site, so you will need to sign in to view them. Thanks, Sam
Nothing in it looks careless. It refers to a plausible meeting, uses a normal tone and asks for a routine action. The risk sits in the sign-in link, because it may lead to a page designed to capture a password.
New warning signs to check
Since the writing no longer gives it away, look at other details:
- The sender address. Check the actual email address, not just the display name. A familiar name can sit on top of an unfamiliar address.
- A link to sign in. Be careful with any unexpected request to log in to view files. Go to the file service directly instead of using the link.
- Any money or bank change. A new payee or changed account details should be confirmed by phone, using a number you already have.
The phishing and IT security checklist covers the controls that support these habits.
Why fewer, better emails cost firms more
Attackers used to rely on volume. They sent huge numbers of poor messages and counted on a few replies. AI tools let them write fewer messages that are better targeted, and those tend to work more often.
The FBI’s Internet Crime Complaint Center reported that business email compromise was one of its largest loss categories in its 2025 report. The figures are reported losses, meaning complaints filed with the FBI, so they do not capture every loss. Check the IC3 annual report for the current numbers before relying on them. Some industry reports also suggest dollar losses have risen faster than complaint counts, which would be consistent with more targeted attacks.
Real-world example
Consider a hypothetical 10-person architecture firm. A project manager receives an email that appears to come from a general contractor she works with. It mentions the correct project and asks her to sign in to review revised drawings.
She notices the sender address ends in a slightly different domain than the contractor normally uses. She does not click the link. She calls the contractor on a number from her contact list, and he confirms he sent nothing.
Her habit of checking the address stopped the problem. Without it, the sign-in page could have captured her email password, and an attacker could have used the account to send payment change requests to clients.