Credential phishing
Credential phishing is an email that poses as a trusted sender and leads the target to a fake sign-in page, where the target types a username and password. The attacker collects them and signs in as the target. Current kits also capture the one-time code, so a code alone does not stop it.
Updated
Phishing and spoofing was the most reported crime type in the FBI IC3 2025 report, with 191,561 complaints. CISA says that more than 90% of successful cyberattacks start with a phishing email.
How credential phishing works
Credential phishing works by moving the target from a message to a page the attacker controls.
- Build the page. The attacker copies a service's sign-in page and hosts it on a lookalike domain.
- Send the lure. An email gives a reason to sign in now: an expiring password, a shared file, a blocked payment.
- Hide the destination. The link text shows a trusted address and the link opens another. See deceptive links.
- Collect the sign-in. The target types a username and password into the copy.
- Use the account. The attacker signs in to the real service, reads mail and may start a business email compromise.
What credential phishing looks like now
Credential phishing now looks clean. CISA notes that poor grammar used to be a common sign, and that in the era of AI some emails have perfect grammar. In this fictional example, the sender poses as Trotta's IT team.
From: Trotta IT <helpdesk@trotta-it.example>
To: Megan Park <megan@trotta.io>
Subject: Password expires today
Your Trotta password expires today.
Sign in to keep your account: trotta.io/ sign-in
[the link opens a different site that copies the sign-in page]Some messages carry the sign-in form in the email itself. Others put the link in a QR code, so the target opens it on a phone.
Signs of credential phishing
| Sign | What it means |
|---|---|
| Link text and destination differ | The visible address is a label. The real one is where the click goes. |
| A sign-in page on an unfamiliar domain | The page is a copy. A password manager will not offer to fill it. |
| A deadline: expires today, account locked | Urgency is the lure. |
| A sender domain that is a near match | A lookalike was registered for the campaign. |
| A sign-in form inside the email | What is typed goes to the sender of the email. |
| A code request the target did not start | Someone else is signing in and needs the code. |
Why one-time codes get stolen too
One-time codes get stolen because the fake page asks for them. The target types the password and then the code into the copy, and the attacker uses both on the real sign-in page while the code is still valid.
The phishing guidance from CISA, NSA, FBI and MS-ISAC lists codes sent by text message, and push prompts without number matching, as weak forms of multi-factor authentication. The guidance recommends phishing-resistant methods such as FIDO. See verification code theft.
Why credential phishing gets through ordinary filtering
Credential phishing gets through ordinary filtering when the link is new and the page behind it is not yet on any blocklist. Attackers change domains often.
Gmail and Outlook screen for phishing before delivery and stop a great deal of it. The messages that pass are the fresh, well-made ones, and those tend to be flagged after delivery.
When an AI agent reads the mail
An AI agent can be phished for credentials in two ways. An agent that browses may follow the link and type a password it holds into the copy.
An agent with access to a human's mailbox can also be used to take accounts over. If the agent can read a reset link or a sign-in code, whoever steers the agent can use it.
How to stop credential phishing
Credential phishing is stopped by making a typed password useless to a thief, and by keeping the lure from being read.
- Use phishing-resistant sign-in. Passkeys and security keys do not work on a copy of the site.
- Go to the site directly. Type the address or use a bookmark instead of the link.
- Reset after a mistake. If a password was typed into a copy, change it and end every session.
- Authenticate your domain. SPF, DKIM and DMARC stop mail that fakes your exact domain.
At the door, before delivery
A pre-delivery inbox checks where links go while the sending server is still connected. Trotta Mail screens for deceptive links and lookalike domains. In the sign-up and verification use case, a fake sign-in message is held while the genuine code is delivered.
For an agent, links that are not clean are defanged in the safe view, so they cannot be followed by accident. Sign-in codes and reset links in a human's mailbox are sealed: to an API key, that message does not exist. See the security model.
No filter catches everything. A well-made page on a fresh domain can still arrive, which is why uncertain mail is held or flagged for a human.
Questions
What are examples of phishing emails?
Common lures are a password that expires today, a shared document, a failed payment and a security alert. Each one ends at a page that asks the target to sign in.
Can phishing get past two-factor authentication?
Yes, when the second factor is a code. A fake page can ask for the code and use it on the real site at once. Passkeys and security keys resist this.
How can I tell a fake login page from a real one?
Check the domain in the address bar, not the look of the page. A password manager that does not offer to fill the password is a sign that the domain is wrong.
What should I do if I typed my password into a fake page?
Change the password at once and sign out of every session. Then check the mailbox for new forwarding rules and tell your security team.