The main types of phishing are email phishing, spear phishing, whaling, smishing, vishing, quishing, clone phishing, angler phishing, and search engine phishing. Each targets a different channel or victim profile, and as of 2026 the taxonomy includes adversary-in-the-middle (AiTM) frameworks and AI-generated attacks.
Phishing remains the entry point for the majority of breaches precisely because it targets people, not systems. The Anti-Phishing Working Group (APWG) tracks tens of thousands of unique phishing sites every month, and the taxonomy keeps expanding as attackers shift channels faster than security teams update their playbooks. This guide maps the full current taxonomy, groups attacks by channel, and gives your analysts the signal characteristics they need to catch each variant before it converts.
Email-Based Phishing: The Channel That Started It All
Email phishing is a social engineering attack where an adversary sends a deceptive message over email to trick the recipient into revealing credentials, transferring funds, or executing a malicious payload. The canonical scenario: a finance analyst receives what looks like a DocuSign request from an unfamiliar sender, clicks through to a credential-harvesting page styled to match Microsoft 365, and enters their password. The attacker now owns that account.
What distinguishes modern email phishing from the campaigns of fifteen years ago is the fidelity. Legitimate mail infrastructure, verified sending domains acquired through lookalike registration or business email compromise, and HTML templates pixel-matched to real service providers make sender-based filtering insufficient on its own. The structural tell of poor grammar has been eliminated by generative AI, which produces grammatically correct, contextually appropriate lures at scale.
Spear Phishing: Targeted, Researched, Harder to Spot
Spear phishing is a targeted email attack crafted around specific intelligence about the victim, typically harvested from LinkedIn, company websites, or prior data breaches. Where bulk phishing casts a wide net, spear phishing fires a single, carefully aimed message.
A realistic scenario: an attacker scrapes a company’s org chart, identifies the CFO’s executive assistant, and sends a message appearing to come from the CFO’s personal email asking the assistant to process an urgent wire transfer before a board meeting. The request references a real upcoming event, uses the CFO’s typical sign-off, and includes a plausible reason not to use internal channels. That specificity is why spear phishing has a significantly higher conversion rate than bulk campaigns.
For attacks targeting C-suite executives specifically, the variant is called whaling phishing, where the financial or reputational stakes per successful compromise are exponentially higher. Whaling attacks frequently combine email with phone follow-up, adding a layer of social proof that generic detection rules do not catch.
Clone Phishing: Your Own Email, Weaponised
Clone phishing is an attack where a legitimate email previously received by the target is copied, its links or attachments replaced with malicious versions, and then resent from a spoofed or compromised address. The attacker’s reasoning is simple: the victim has already engaged with this email once, so the message content carries built-in credibility.
Scenario: a vendor sends a routine invoice with a PDF attachment. Hours later the victim receives what appears to be the same email, this time from a slightly different sender domain, with an updated PDF that contains a macro dropper. The subject line, body copy, and attachment name are identical to the original. Without side-by-side comparison, most recipients would not notice the difference.
Clone phishing is particularly effective against finance and procurement teams that process high volumes of repeat vendor communications. Detection depends on header analysis and attachment sandboxing, not content filtering alone.
Phone and SMS Phishing: Smishing, Vishing, and AI Voice Cloning
Smishing is phishing conducted via SMS, where the attacker sends a text message impersonating a trusted entity, such as a bank, courier service, or government body, to prompt the recipient to click a link or call a number. Vishing is voice phishing conducted by phone call, where the attacker impersonates technical support, law enforcement, or a known colleague to extract credentials or authorise actions verbally.
Smishing scenario: a warehouse operations manager receives a text claiming their Royal Mail parcel requires a customs payment of £2.99 before delivery. The link resolves to a convincing Royal Mail clone that harvests card details. The £2.99 figure is psychologically calibrated to be too small to trigger scepticism, while the card data captured has substantially higher value.
Vishing has been meaningfully changed by AI voice cloning as of 2026. Adversaries now generate synthetic audio from as little as three seconds of source material, which is trivially available from public LinkedIn video posts or recorded earnings calls for publicly traded companies. The cloned voice calls a target’s colleague claiming to be the executive, authorises a transaction verbally, and hangs up before any secondary verification check is considered. The UK National Cyber Security Centre has published guidance on AI-enabled impersonation risks, noting that verification protocols must no longer rely on voice recognition alone.
I reviewed a vishing incident response report from a financial services client in late 2025 where the threat actor used cloned audio of the CISO to instruct a junior IT admin to reset MFA for a privileged account. The admin followed the instruction because the voice was indistinguishable from the real person. The breach was detected only because the account’s subsequent activity triggered a UEBA alert, not because any human flagged the call as suspicious. That incident clarified how quickly the verification assumptions built into most security awareness programmes have become liabilities.
QR Code Phishing (Quishing) and Search Engine Phishing
Quishing is phishing conducted through malicious QR codes, where the attacker embeds a URL in a QR image that redirects the scanner to a credential-harvesting page or initiates a malicious download. Because the destination URL is not visible before scanning, most email security gateways that inspect hyperlinks do not analyse QR code payloads by default.
Quishing emerged as a significant attack vector in 2023 and has matured considerably as of 2026. Attackers embed QR codes in PDF attachments, printed materials left in office common areas, or overlaid on legitimate public QR codes such as restaurant menus and parking payment terminals. The mobile device used to scan typically has weaker email security controls than a corporate laptop, creating a gap in the protection chain.
Scenario: an employee receives a phishing email warning that their Microsoft Authenticator needs reconfiguration. The email body contains no URL, only a QR code. The employee scans it on their personal phone. The resulting page is a convincing Microsoft login portal that captures credentials and, via an AiTM proxy, also captures the session token issued after MFA completion, bypassing two-factor authentication entirely.
Search engine phishing, sometimes called SEO poisoning, is a technique where attackers create convincing malicious websites and optimise them to appear in search results for brand-name queries or financial service searches. The victim does not click a link in an email; they open a browser, search for their bank’s login page, and click what appears to be the correct result. The Cybersecurity and Infrastructure Security Agency (CISA) has documented SEO poisoning campaigns targeting taxpayers during peak filing periods, where malicious sites mimicking government portals captured both credentials and Social Security numbers.
Social Media Phishing: Angler Attacks in Real Time
Angler phishing is a social media attack where adversaries monitor brand mentions on platforms such as X, LinkedIn, or Facebook and respond to customer complaints or queries from fake brand support accounts. The attacker intercepts a conversation in progress, offering to help resolve the issue, and directs the victim to a phishing page under the pretence of account verification or ticket creation.
Scenario: a retail customer tweets a complaint about a delayed order to a major supermarket chain. Within minutes, a fake account with a similar handle and the brand’s logo responds, offering to help via a direct message link. The link routes to a form requesting the customer’s account email, password, and payment card number to locate the order. The attacker monitors high-volume brand mentions in real time, making this attack both scalable and highly contextual.
Angler phishing is difficult to detect at the organisational level because the initial lure happens outside the corporate perimeter. Brand protection monitoring tools can alert on impersonating accounts, but response latency is the critical variable: the attack completes in minutes, often before any takedown request is processed.
MFA Fatigue and Adversary-in-the-Middle: The 2026 Escalation
Two attack patterns define the 2026 phishing escalation. MFA fatigue is the standard follow-on to a successful credential-phishing step: the attacker takes stolen credentials and repeatedly triggers push notifications to the victim’s authenticator app, often overnight, until the victim approves one to stop the interruption. The counter-control is number matching with additional context display, which Microsoft Authenticator and Duo both support. Organisations still using simple approve or deny push notifications without context remain exposed. Adversary-in-the-middle (AiTM) phishing goes further: a real-time proxy relays the victim’s authentication session, capturing both credentials and the session cookie issued after MFA completes. Frameworks such as Evilginx and Modlishka automate this. The victim completes a genuine MFA challenge; the attacker captures the session token and uses it before expiry. The only controls that hold against AiTM are FIDO2/passkey authentication and phishing-resistant MFA bound to the origin domain, because session binding prevents token replay to a different origin.
This intersection of phishing and identity security means that your cloud security best practices, particularly token lifetime policies and conditional access controls for Microsoft 365, Google Workspace, and AWS IAM, are directly load-bearing in your phishing defence architecture, not a separate conversation.
Phishing Attack Taxonomy: Type, Channel, Target, Signal
The table below gives your SOC analysts a quick reference across all major variants currently active as of 2026.
| Attack Type | Primary Channel | Typical Target Profile | Key Detection Signal |
|---|---|---|---|
| Bulk email phishing | All employees | Lookalike sender domain, generic salutation | |
| Spear phishing | Named individuals, finance, HR | Correct name/role, reference to real internal context | |
| Whaling | Email and phone | C-suite, board members | High-urgency financial request, out-of-band confirmation bypass |
| Clone phishing | Finance, procurement | Resent legitimate message, sender domain mismatch on inspection | |
| Smishing | SMS | Consumers, remote workers | Delivery/payment urgency, short URL, no prior relationship context |
| Vishing (incl. AI voice) | Phone call | IT admins, finance teams | Authority pressure, request outside normal process channel |
| Quishing | QR code in email or print | Office staff, remote workers | QR in email body, no accompanying URL text, mobile-only flow |
| Search engine phishing | Search results | All users, peak event periods | Paid ad or result for brand name, URL does not match official domain |
| Angler phishing | Social media | Customers in service conversations | Unsolicited support response, similar-but-wrong account handle |
| AiTM phishing | Email and web proxy | Any MFA-enabled account | Sign-in from unfamiliar IP immediately post-authentication, session anomaly |
How AI Has Changed Phishing Attacks in 2026
Generative AI has removed the two most reliable low-effort detection signals that security awareness training relied on for years: poor grammar and generic content. AI-written phishing emails are grammatically correct, contextually appropriate, and can be generated at scale for hundreds of individual targets simultaneously, each personalised with details that previously required hours of manual open-source intelligence gathering.
The more significant shift is speed. AI models can generate a convincing phishing campaign, complete with domain registration guidance, email infrastructure setup, and personalised lure content, in a fraction of the time manual adversaries required. This compresses the window between credential theft and account takeover to a degree that automated detection, not human review, must be the primary control layer.
AI voice cloning for vishing, AI-generated deepfake video for executive impersonation on video calls, and LLM-assisted spear phishing that synthesises months of a target’s public communications into a personalised lure are all documented in incident reports as of 2026. Your detection and response playbooks need to account for the fact that content quality is no longer a reliable indicator of legitimacy.
Organisations that have reduced phishing susceptibility in this environment share one characteristic: they have moved from awareness-first programmes toward control-first architectures. Phishing-resistant MFA, email authentication enforcement at DMARC reject policy, and automated QR code inspection in email gateways do more measurable work than any volume of simulated click-rate training metrics.
Phishing Attack Types: Common Questions Answered
What are the main types of phishing?
The main types of phishing are email phishing, spear phishing, whaling, clone phishing, smishing (SMS), vishing (voice), quishing (QR code), angler phishing (social media), and search engine phishing. As of 2026, adversary-in-the-middle and AI-assisted variants have become significant sub-categories across several of these channels.
What is quishing?
Quishing is phishing delivered through a malicious QR code embedded in an email, printed document, or physical object. When scanned, the QR code redirects the victim to a credential-harvesting page. Most email security gateways do not inspect QR payloads by default, which is why quishing bypasses conventional link-scanning controls.
What is the difference between smishing and vishing?
Smishing uses SMS text messages as the attack channel, typically delivering a malicious link or a call-back number. Vishing uses live or recorded phone calls to impersonate authority figures and extract credentials or authorise actions verbally. Both exploit trust signals that email security controls do not cover, making them effective against targets with strong email defences.
How has AI changed phishing attacks?
AI has eliminated grammar and personalisation as reliable detection signals. Attackers now generate correctly written, individually tailored lures at scale in minutes. AI voice cloning enables convincing audio impersonation from minimal source audio. Adversary-in-the-middle frameworks automate session token capture post-MFA. Content quality can no longer be used as a primary indicator of a phishing attempt.
