By Derek Manky, Chief Security Strategist & Global VP Threat Intelligence at FortiGuard Labs
For years, phishing awareness centered on identifying obvious signs like misspelled words and awkward phrasing. Although these indicators still exist, they no longer serve as dependable screening tools. AI-created messages can be refined, personalized, and aligned with a person’s or organization’s usual language.
The more useful question is no longer, “Does this message look professional?” It is, “Does this request make sense, and have I verified who is making it?”
This change matters because an attacker only needs their victim to perform one action, such as opening a file, scanning a QR code, revealing sensitive information, or transferring money. As a result, protecting against phishing in the AI era should focus less on superficial message flaws and more on the actions the message aims to trigger.
The Message Is Only the Delivery Mechanism
Phishing remains effective because it provides an entry point to multiple forms of compromise. A credential-harvesting page can capture a username and password. A malicious attachment can trigger a malware infection. A fraudulent payment request can redirect funds. And a compromised account can then be used to send more convincing messages from a trusted address.
The phishing message is often the least technically complex part of an attack. Its purpose is to persuade the recipient to cross a trust boundary by entering credentials, approving an authentication request, opening a file, or changing an established process. Once that happens, the attacker may steal credentials or session tokens, install malware, establish persistence, or use the compromised account to target others. Adding AI has not fundamentally changed those objectives. Instead, it has improved the quality, personalization, and scalability of the lure, increasing the probability that an otherwise familiar attack chain will succeed.
Attackers also often create campaigns tied to events already attracting public interest. For example, FortiGuard Labs found over 13,000 domains related to FIFA World Cup 2026 registered between January and May, with about 8.8% identified as malicious or suspicious. These include fake tickets, job ads, streaming links, and impersonation accounts, all effective because they appear relevant and credible in the current context. AI enhancements helped make these domains look legitimate while removing many of the tells that identify a phishing message or site.
AI Changes the Economics of Social Engineering
AI has also made spear phishing more effective. Effective spear phishing traditionally required time, with an attacker needing to research a target, understand the target’s role and relationships, and then construct a message appropriate to that context. AI reduces the effort required at each step.
AI capabilities can support convincing images, audio, and video impersonations. Of course, these techniques do not make every phishing attempt sophisticated, nor do they guarantee success. They do, however, make familiar signals such as polished grammar, an appropriate tone, or a recognizable voice less useful as evidence of authenticity.
This is why awareness programs must evolve from detecting poorly written messages to validating identity, context, and intent.
Verify the Request, Not Just the Message
Employees should exercise caution and slow down if they receive a communication asking for credentials, money, sensitive data, MFA approval, or urgent action, particularly if the request is unusual or deviates from standard procedures.
Verification should be carried out independently. Avoid using the phone number, reply address, or link from the suspicious message. Instead, contact the person through a trusted number, corporate directory, or official internal channel. For any payment or account changes, follow the organization’s approval procedures, even if the request appears to come from an executive or a trusted supplier.
Employees should also consider the context surrounding the request. Was the communication expected? Is this how the person normally makes this type of request? Does the destination domain match the organization it claims to represent? Is the message asking the recipient to avoid normal review or keep the action confidential?
No single answer can prove that a message is malicious. Together, however, these checks can reveal inconsistencies that polished language and realistic branding are designed to conceal.
Train for Decisions, Not Just Recognition
Organizations need to train employees to recognize and respond to manipulation attempts by attackers. Training should include email, text messages, collaboration tools, QR codes, phone calls, and new types of synthetic media. It should also focus on roles and business processes that are most vulnerable to targeted attacks.
Phishing simulations provide a controlled way to measure how users respond to realistic scenarios and pinpoint where additional training or security controls are needed. FortiSAT, Fortinet’s unified security awareness and phishing simulation platform, which incorporates capabilities previously delivered through FortiPhish, uses realistic phishing campaigns and user-risk data to identify risky behaviors, calculate individual risk scores, and deliver targeted remediation. These risk scores can also be leveraged by other Fortinet security products to dynamically enforce tighter security controls for high-risk users, enabling a more adaptive, risk-based approach to security.
The objective is not to catch employees making mistakes. It is to give them repeated practice in recognizing when a request crosses a trust boundary and in knowing what to do next.
Phishing is evolving, but the most effective response is straightforward: Pause. Verify. Report. Treat unexpected requests for credentials, money, sensitive information, or urgent action with caution, and independently verify them before proceeding.