- 01The signs of a phishing email are in specific places
- 02The sender and the link can be checked without opening anything
- 03Email authentication checks the exact domain and does not cover lookalikes
- 04The signs fail when the email fits your work
- 05Before and after the click, the step is to verify through another channel and report
- 06What VU does
- NIST’s Phish Scale (Technical Note TN 2276, November 2023) brings together 23 signs of a phishing email in five categories: errors, technical indicators, visual presentation, language and content, and common tactics.
- DMARC, an IETF standard since May 2026 (RFC 9989), does not cover visually similar domains or the sender’s display name, according to its own section 2.2.
- The FBI warned in December 2024 that generative AI can correct the human errors that previously served as a sign of fraud.
- If you handed over your password, CERT.br (2026) says to change it in every account where you use it and close active sessions on other devices.
The email that the United States Federal Trade Commission (FTC) uses to teach people how to recognize phishing has the correct company logo in the header. What gives it away is elsewhere: a generic greeting, a notice that the account is on hold because of a billing problem, and a link to update payment details. According to the FTC, legitimate companies do not ask you to update that information through an email link.
CERT.br, Brazil’s incident response center, defines phishing as fraud that seeks to obtain personal and financial information by combining social engineering with technical resources. By email, that takes the form of a message impersonating a company, an agency, or someone known to you so that you hand over data, click a link, or open an attachment. The general definition and the variants through other channels are in phishing and smishing.
Its signs have specific places within the email. One part is normally checked by your email provider before showing you the message, another part remains on your side, and there are well-crafted emails that show almost none at all.
The signs of a phishing email are in specific places
NIST, the U.S. standards institute, groups 23 signs of a phishing email in its Phish Scale into five categories, and says where in the email each one appears:
- Errors — spelling and grammar mistakes, and inconsistencies, such as a signature that does not match the sender or an attachment that differs from the one mentioned in the text. They can appear anywhere in the email.
- Technical indicators — the sender address, links, and attachments. Also anywhere.
- Visual presentation — the brand, logos, design, and format. They are usually in the body of the message.
- Language and content — the generic greeting, the lack of details about the signer, the request for sensitive information, urgency, and threatening language. They are in the subject and in the message.
- Common tactics — offers that are too good or available for a limited time, humanitarian requests, imitating a work process, and impersonating a friend, a colleague, a boss, or an authority. Also in the subject and in the message.
NIST counts them for a practical reason: an email with fewer signs is harder to detect, and one with more signs is easier.
CERT.br reaches the same tone-related signs in its 2026 guide, written in Portuguese: threat, haste, emotional appeal, a too-good advantage, obedience to authority. And it proposes three questions before acting: whether the message says it comes from someone known to you, whether it involves money, benefits, or personal data, and whether it wants you to do something without time to think.
The sender and the link can be checked without opening anything
Three of the Phish Scale’s technical signs can be checked by looking, without opening attachments or clicking.
The first is the sender. NIST counts it when the display name does not match the address in the "From" field or the reply-to address: its example is a name such as "IT Helpdesk" with a reply-to address such as accounts-payable@gmail.com.
The second is the link, whose visible text can hide the real URL; on a computer, the URL appears when you hover over it before clicking. The third is the lookalike domain, an address or link that imitates a known site, such as the amazan.com used as an example by the United States Cybersecurity and Infrastructure Security Agency (CISA).
Gmail and Outlook add two marks next to the sender. A question mark indicates that the provider could not authenticate the message: Gmail does not know whether it comes from who it appears to come from. The word "via" next to the name indicates that the domain from which it was sent does not match the sender that is displayed. Microsoft adds a clue: if a sender you know never had the question mark and it suddenly appears, they may be being impersonated.
Neither of the two marks is enough to decide: Google clarifies that an unauthenticated message is not necessarily spam, as happens with some mail sent to mailing lists, and Microsoft says the same about messages with "via." In a corporate Outlook account, moreover, they depend on the anti-phishing policy configured by the organization, so they may not appear.
Email authentication checks the exact domain and does not cover lookalikes
Email authentication checks whether the sender’s domain is the one it claims to be, but it does not detect a lookalike domain registered by the attacker. Your email provider normally checks it before showing you the message, with three IETF standards:
- SPF (RFC 7208) — authorizes which servers can send on behalf of a domain, but for the SMTP envelope address, not for the sender you see.
- DKIM (RFC 6376) — signs the message with a domain that takes responsibility for it. A valid signature says which domain signed, not that the apparent author is who they claim to be.
- DMARC (RFC 9989) — ties the result to the visible sender’s domain, because it requires it to match an already authenticated identifier. The domain owner publishes what to do if it fails: no action, quarantine, or rejection. It has been an IETF standard since May 2026 and replaces RFC 7489 from 2015.
The joint guidance from CISA, the NSA, the FBI, and the MS-ISAC from October 2023 recommends enabling DMARC for received email and configuring it to reject for sent email.
The limit is written by the DMARC RFC itself, in its section 2.2: the standard "does not address the use of visually similar domain names or abuse of the RFC5322.From human-readable display name". It does not cover lookalike domains or the sender’s display name.
A consequence follows from that, which is our reading of the three RFCs and Microsoft’s documentation, and not a sentence from any of them: an attacker who registers a lookalike domain and configures it with their own SPF, DKIM, and DMARC passes all three checks, because they are authenticated for their domain, and also shows neither a question mark nor "via."
Google says it another way: "spammers can also authenticate mail". The lookalike domain and the display name remain on the reader’s side, and they are precisely two of the technical signs from the previous section.
The signs fail when the email fits your work
A list of signs works worse with better-crafted emails. NIST measures this with a second axis of the Phish Scale, premise alignment: how closely the email resembles the tasks and responsibilities of the person who receives it. The more it resembles them, the harder it is to detect.
An email with few signs and a strong premise, such as one that imitates a routine company process at the time when that process occurs, is, according to NIST, very difficult to detect.
The original Phish Scale paper, published by NIST researchers in 2019, gives the reason: "The greater the contextual relevance, the less likely a user is to notice, attend to, and think deeply about suspicious email cues." When the premise targets a specific person, NIST brings it closer to spear phishing. Why the deception works even with people who have been warned is covered in social engineering.
Spelling used to be the classic sign, and it is losing value: the FBI warned in December 2024 that generative AI tools "can correct for human errors that might otherwise serve as warning signs of fraud". The warning talks about fraud in general and does not measure how many phishing emails now arrive without errors. An email with mistakes is still suspicious, and one without mistakes proves nothing.
The fact that the email includes real data about you does not make it legitimate either. CERT.br’s 2026 guide warns that scammers use leaked and public information to make the scheme look real.
NIST itself closes its guide with that limit: awareness is not a magic solution, and an organization needs to combine technology, processes, and people to identify, respond to, and report suspicious emails.
Before and after the click, the step is to verify through another channel and report
Before the click, the step is to verify the message through another channel and report it; afterward, close off whatever was handed over.
When faced with a suspicious email, CISA recommends not replying or clicking any link or attachment, including the unsubscribe link. If the message could be real, the contact is looked up through another route. The FTC suggests asking yourself whether you have an account with that company and, if you do, contacting it through a phone number or website that you know is real, not with the details in the email. Verifying through another channel has its own development.
Reporting the email protects others. Gmail and Outlook have the option to report phishing in the menu of the message itself, and the impersonated company is notified through its official channel, never by replying to the email. CISA’s guide counts reporting among the most efficient methods to protect an organization, because it helps email providers identify new campaigns.
If you already clicked or handed over data, CERT.br organizes the steps in its booklet Caiu? Veja o que Fazer (2026):
- If you handed over your password — change it immediately in every account where you use it and close active sessions on other devices.
- If you had two-factor authentication — reject approval notifications that you did not initiate.
- If you opened a file or link on the computer — run a full antivirus scan, delete suspicious files, and reinstall the system if you cannot remove the malware.
- In all cases — save everything you have about the scam: links, messages, screenshots, and receipts. In Brazil, CERT.br also says to file a police report.
Why closing sessions matters as much as changing the password can be seen in the attack step by step.
If the account was a work account, the organization has its own steps. The joint CISA guidance says to re-provision compromised accounts, audit their access after the incident, and isolate the affected device.
What VU does
The VU capability that provides context about the device is Protect against fraud, with recognition of the device from which the operation is performed based on browser and mobile signals. The decision about what to do with that signal, and monitoring the operation, are resolved in another layer that VU does not provide.
Before the click, it is up to you to look at what your email client does not flag: the lookalike domain, the display name, and a premise that fits too well.
schedule a conversation.
