What changed
Phishing used to announce itself with broken English and a Nigerian prince. Now it is written by the same language models that write everything else, uses the real company’s templates and logos, and arrives seconds after you actually did something, like buying a flight or resetting a password, because the sender bought data about you. The tells have moved from the text to the structure of the request.
The tells that still work
1. It wants you to act now
“Your account will be closed in 24 hours.” “Unusual sign-in, confirm it is you.” “Your parcel is held pending a small fee.” Urgency exists to stop you thinking. Real companies rarely need you to do anything within the hour, and never through a link in an unexpected message.
2. The link does not go where it says
Hover over the link (or long-press on a phone) and read the actual destination. The domain is the part just before the first single slash, read from the right: in https://paypal.com.secure-login.info/verify the domain is secure-login.info, not PayPal. Look-alikes use rn for m, 1 for l, extra words (paypal-support.com), or a legitimate hosting service with the brand in the path. Long tracking links can be pasted into the URL decoder to see their real parts.
3. It asks for something the real service never asks for
Your password, your 2FA code, your full card number, a gift card. No bank, no tech company and no government agency asks for a password or a one-time code by email, text or phone. If a “support agent” asks you to read out the code that just arrived, that code is being used to log in as you right now.
4. The sender is almost right
Display names are free text; anyone can be “Apple Support”. Open the actual address. [email protected] is Apple; [email protected] or [email protected] is not. On a phone, tap the name to see the address.
5. It arrived through an unusual channel
A text from your bank with a link, a WhatsApp from your boss asking for gift cards, a LinkedIn message with a document to open. Real organisations are boring and predictable about how they contact you. Novelty is a signal.
The habit that beats all of it
Never log in from a link in a message. If the message says there is a problem with your account, close it, open a new tab, type the address you know (or use your bookmark or the app), and log in. If there really is a notification, it will be there. This single rule defeats every phishing email ever sent, including the ones you cannot spot, because you never hand your credentials to a page you did not navigate to yourself.
Password managers enforce this automatically: they fill credentials only on the exact domain they were saved for. If the manager does not offer to fill your bank password, you are not on your bank’s site. Treat that silence as an alarm.
When you already clicked
- If you typed a password: change it on the real site immediately, then everywhere you reused it. Check it with the breach checker once it is retired.
- If you typed a 2FA code: the attacker likely has a session. Log in on the real site, sign out all sessions, check recovery settings for additions.
- If you entered card details: call the bank.
- If you downloaded or ran something: disconnect, run a full scan, and change passwords from a different device.
Our breach checklist covers the same steps in order.
Making phishing fail even when it works
Codes can be relayed; signatures cannot. Passkeys and hardware keys sign a challenge that names the real domain, so a login attempt on a fake site produces nothing an attacker can use. Where a service offers either, turn it on. Then the question of whether you can spot the email stops mattering.