How to Tell If an Email or Message Was Written by AI

A well written email used to be a good sign. Bad spelling and clumsy sentences were often the easiest way to spot a scam. That is changing fast.

AI writing tools can now produce a warm, polished, error free message in seconds. Scammers have noticed, and so have honest businesses. The result is that both real and fake messages can look equally professional.

This guide walks through what AI-written text actually looks like, which old warning signs no longer work the way they used to, and what to check instead. None of it requires special software or technical skill, just a few minutes and a checklist.

How Can You Tell If an Email or Message Was Written by AI?

You usually cannot tell for certain just by reading it, so focus on what the message is asking you to do rather than how well it is written. AI text has a few common habits, but none of them prove anything on their own. The stronger clues are in the request itself: is it urgent, secretive, or asking for money, passwords, or remote access to your computer.

That said, AI-written messages do tend to share a few patterns worth knowing.

Overly formal or oddly smooth tone

AI tools often default to a polite, formal register, even in a message meant to feel personal. Phrases like “I hope this message finds you well” or “please do not hesitate to reach out” appear constantly, because they are common in the training text AI learned from. A real family member or a real company representative writing quickly is more likely to sound a little rougher around the edges.

Generic, could-apply-to-anyone phrasing

AI cannot see your actual account, your actual order, or your actual family. So an AI-drafted scam message tends to stay vague: “your recent order,” “your account,” “a family member,” rather than your real order number, the real product you bought, or your grandchild’s real name. A message that avoids specifics while sounding confident is worth a second look.

Unusual perfection, with no personality

Real people writing quickly leave small fingerprints: a dropped word, an odd abbreviation, a sentence that trails off, a joke, a typo. AI text is often smooth from the first word to the last, with balanced sentences and tidy paragraphs. That polish is not proof of a scam, plenty of real businesses now use AI to draft their emails too, but it does mean you cannot rely on “it sounds too professional to be a scam” anymore.

Repeated structure and safe, hedged language

AI often falls back on lists, “on one hand, on the other hand” style balance, and cautious wording that avoids saying anything too specific or too risky. A message that reads like it was written to satisfy nobody in particular, rather than to talk to you directly, can be an AI hint.

Why the old “bad grammar means scam” rule stopped working

For years, one of the most reliable scam clues was clumsy English: missing articles, odd word order, or spelling mistakes, especially from scammers operating overseas. That rule is fading.

The FBI’s Internet Crime Complaint Center (IC3) has warned that criminals now use generative AI tools to translate and polish their messages, cutting down on the grammar and spelling errors that used to give them away. The same alert notes that AI lets scammers write convincing messages faster and send them to more people at once.

That does not mean grammar checking is useless, a message riddled with strange errors is still worth extra suspicion, it means good grammar can no longer be your main reason to trust something.

A quick before and after: what changed

Steps to check a message you are unsure about

1. Read the request, not the writing. What is the message actually asking you to do? Click a link, call a number, send money, share a code, or give remote access to your computer are all reasons to slow down.
2. Look for real, specific details. A genuine message from your bank, a delivery company, or a family member usually includes something specific to you: an actual order number, an actual amount, a detail only they would know. Vague, generic wording is a caution sign.
3. Check who it is really from. Hover over or tap the sender’s name to see the full email address, not just the display name. Compare it letter by letter with a known, real address, not one copied from the suspicious message itself.
4. Slow down on urgency. Messages that pressure you to act “immediately” or “before your account is suspended” are designed to stop you from thinking clearly. AI can make that pressure sound calm and reasonable instead of panicked, which makes it easier to miss.
5. Verify through a separate channel. Call the person or company using a phone number you already have, from a past bill or the back of your card, never a number provided in the message itself.
6. Ask directly. If it is from a real person you know, it is completely fair to ask, “Did you write this yourself?” A genuine sender will not be offended.

Can an AI tool tell you if text was AI-written?

There are online tools that try to guess whether a piece of text was written by AI, by looking at word choice and sentence patterns. They can be a helpful extra check, but they are not reliable enough to trust on their own. They can wrongly flag genuine human writing as AI, and they can miss AI text that has been edited by a person afterward. Treat any such result as one more data point, not a verdict.

You can also ask an AI assistant you already trust, like the one on your phone or computer, to help you think through a suspicious message.

AI writing does not automatically mean a scam

It is worth remembering that plenty of legitimate messages are now AI-assisted too. Customer service replies, marketing emails, and even some replies from real companies may be drafted or polished with AI. According to Pew Research Center, roughly a third of U.S. adults had used ChatGPT by early 2025, and that number keeps climbing, which means AI-assisted writing is becoming an ordinary part of everyday communication, not just a scam tool.

So the goal is not to reject every polished message. It is to stop using “it sounds professional” as your main safety check, and start checking the request itself, the sender, and the specific details instead.

When to loop in someone else

If you are ever unsure, it is always fine to pause and ask someone you trust before responding, clicking, or sending anything. A family member, a friend, or even a call to your bank’s official number can settle the question in a few minutes. There is no such thing as too cautious when money, passwords, or personal information are involved.

Our Scam-Stopper Kit (free when you join our email letter) includes a simple checklist you can keep by your computer for exactly this kind of moment.

You may also want to read about how to tell if a package delivery text is a scam, how romance scams use AI to build trust, and where to report a scam once you have spotted one. If you want to understand the AI tools themselves a little better first, is ChatGPT safe and free is a good place to start, and if voice calls worry you as much as messages do, see how to set up a family safe word against AI voice-clone scams.

Where to go from here

Recognizing an AI-written message is less about spotting perfect grammar and more about slowing down, checking specific details, and verifying through a channel you already trust. For a broader, plain-English guide to living alongside AI safely, from spotting scams to using AI tools yourself with confidence, our book AI for Seniors covers this in more depth.