Phishing statistics are unusually easy to get wrong, because the headline numbers come from sources that count completely different things.
The FBI publishes complaints, which are reports from victims. APWG publishes observed attacks, which are reported phishing sites and campaigns. Microsoft publishes detections, which are threats its systems saw in email. None of those are the same measurement, and adding them together produces a number that means nothing.
This page keeps them apart. Every figure below says who counted it, what they counted and when.
Phishing Statistics 2026: Editor’s Choice
- The FBI received 191,561 phishing and spoofing complaints in 2025, making it the most-reported crime type. (FBI IC3)
- Those complaints reported $215,843,126 in losses.
- Business email compromise was separate: 24,768 complaints and $3.05 billion in losses.
- APWG observed 3.8 million phishing attacks during 2025, up only slightly from 3.76 million in 2024.
- Attacks rose 10.1% in Q2 2026, from 971,181 to 1,069,681, with 425,808 in June alone. (APWG)
- Smishing rose 40% quarter on quarter, and vishing 20%.
- Microsoft detected about 8.3 billion email-based phishing threats in Q1 2026, falling to 7.6 billion in Q2. (Microsoft)
- QR code phishing rose 146%, from 7.6 million in January 2026 to 18.7 million in March.
- In a cited academic study, AI-automated phishing achieved 4.5 times the click-through rate of standard attempts.
How Much Phishing Is There?
Three organisations publish credible volume data, and each answers a different question.
Victim reports: 191,561 complaints to the FBI in 2025
The FBI’s Internet Crime Complaint Center logged 191,561 phishing and spoofing complaints in 2025, out of 1,008,597 complaints across all crime types. That made phishing and spoofing the single most-reported crime type, ahead of extortion at 89,129 and investment fraud at 72,984.
Read that as a floor rather than a count of attacks. It records people who were targeted, noticed, and chose to file a report with the FBI. Most phishing never reaches that threshold.
Observed attacks: 3.8 million in 2025
APWG, which collects reports of phishing sites and campaigns from its member companies, observed 3.8 million phishing attacks during 2025. The direction matters as much as the number: that was up only slightly on the 3.76 million seen in 2024.
Quarterly, attacks rose 10.1% in Q2 2026, from 971,181 in Q1 to 1,069,681 in Q2, the highest quarterly total since Q3 2025. June 2026 alone accounted for 425,808 attacks.
Detections: 8.3 billion email threats in a single quarter
Microsoft Threat Intelligence detected approximately 8.3 billion email-based phishing threats in Q1 2026, and about 7.6 billion in Q2, with monthly volumes easing from 2.7 billion in April to 2.4 billion in June.
The gap between 3.8 million and 8.3 billion is not a contradiction. APWG counts reported attacks and campaigns; Microsoft counts individual messages its systems flagged. For scale, Microsoft says it screens about five billion emails a day for malware and phishing, which is a screening figure, not a count of phishing.
Taken together, the honest summary for 2026 is that phishing is enormous, roughly flat in volume, and changing fast in method. That last part is where the interesting data is.
What Phishing Costs
Reported losses from phishing and spoofing complaints came to $215,843,126 in 2025.
Two things to keep straight about that figure. It is an aggregate of what victims reported, not an average cost per attack, and the old claim on this page that a phishing attack costs a mid-sized company $1.6 million has been removed because no current source supports it. It is also only the losses people reported to the FBI, so it understates the real economic cost.
It is also, notably, small next to other categories. Total reported losses across all crime types were $20.877 billion, up 26% on 2024, with investment fraud alone accounting for $8.6 billion. Phishing generates far more complaints than almost anything else while producing comparatively modest direct reported losses, because it is usually the way in rather than the payoff.
That is the key point about phishing economics. The money tends to show up in whatever happens next, which is why business email compromise is counted separately.
Business Email Compromise
BEC is a different fraud category and should never be added to phishing losses. It involves compromising or impersonating a trusted email account to redirect a payment, and it is far more lucrative per incident.
In 2025 the FBI recorded 24,768 BEC complaints and $3,046,598,558 in reported losses. That is roughly an eighth as many complaints as phishing, producing about fourteen times the reported losses.
APWG tracks attempted theft amounts in wire-transfer BEC. In Q2 2026 the average amount requested was $61,732, a 45% increase on the previous quarter’s $42,663, and the number of wire-transfer BEC attacks rose 88%. (APWG)
Note the wording: that is the amount attackers asked for, not the average amount they got. Attempted theft is not realised loss, and treating the two as the same is a common error.
How Attacks Are Changing
Volume is flat. Delivery methods are not, and they move faster than most statistics pages can keep up with.
QR code phishing rose 146% in three months, then changed shape
Microsoft recorded QR code phishing climbing from 7.6 million attacks in January 2026 to 18.7 million in March, a 146% increase and the highest monthly volume in at least a year. QR codes let attackers hide a URL inside an image that text-scanning engines cannot read, and push the victim onto a phone that company security tools may not cover.
Then it changed again. Email-embedded QR codes, which had surged 336% in March, effectively disappeared in Q2 2026, dropping to near zero across all three months as attackers moved back to attachment-based delivery. PDFs were already the dominant carrier, growing from 65% of QR attacks in January to 70% in March.
A tactic going from fastest-growing to near-extinct within six months is the clearest argument against treating any lure ranking as durable.
Smishing rose 40% and vishing 20% in a single quarter
APWG recorded smishing, phishing by SMS, rising 40% from Q1 to Q2 2026, with vishing, phishing by voice call, up 20% to record highs. Email remains the main channel, but it is no longer the only one worth defending.
Telecom became the most-attacked sector
In Q1 2026, APWG’s members found Telecom the most-attacked sector, rising from 5.9% of all attacks in Q3 2025 to 33%, with SaaS/Webmail and social media also heavily targeted. (APWG)
Sector rankings shift quarter to quarter, which is exactly why this page no longer carries the 2021 rankings it used to. Quote them with the quarter and the source attached, or not at all.
For a stage-by-stage view of delivery, credential theft, user interaction, and downstream compromise, see these phishing statistics by attack stage.
AI and Phishing
This is the area where exaggeration is easiest, so it is worth being precise about what has actually been measured.
The most-quoted figure comes from Microsoft’s Digital Defense Report 2025: AI-automated phishing emails achieved 54% click-through rates against 12% for standard attempts, a 4.5 times increase.
Microsoft is citing research, not its own production telemetry. The underlying work is a Harvard Kennedy School paper, published as an arXiv preprint in November 2024, which tested fully automated spear phishing on human subjects. So it is a controlled study on a limited group, not a measurement of what every phishing campaign now achieves, and the sensible way to quote it is as a multiple rather than as a universal click rate.
The same report notes AI automation could increase phishing profitability up to 50 times by making highly targeted attacks cheap at scale. That is a projection, not an observation.
There is one current first-party measurement worth adding. The FBI now tracks an AI-related descriptor on complaints, which appeared on 22,364 complaints in 2025 with $893,346,472 in reported losses. That is across all crime types rather than phishing alone, but it is the closest thing to an official count of AI-assisted fraud reaching victims.
Types of Phishing
The vocabulary is worth keeping straight, because the defences differ.
- Phishing is the broad category: a message impersonating someone trusted to get credentials, payment or access.
- Spear phishing targets a specific person with details that make the message credible. The AI research above is about automating exactly this.
- Whaling is spear phishing aimed at senior executives, usually to authorise a payment.
- Smishing arrives by SMS, vishing by voice call, and quishing through a QR code.
- Business email compromise uses a compromised or spoofed account to redirect a real payment, and is counted separately by the FBI.
The old version of this page called spear phishing the most common type. That ranking was not sourced and has been removed; what the current data supports is that spear phishing is the most consequential, not the most numerous.
Notable Cases, Clearly Dated
These are historical, and included because they illustrate how the mechanics work rather than as evidence of current prevalence.
- Google and Facebook, 2013 to 2015. A Lithuanian national invoiced both companies for goods that did not exist, using a fake supplier identity. They paid roughly $100 million between them before it was discovered.
- Bangladesh Bank, 2016. Attackers used phishing as the initial access vector in a chain that moved $81 million out of the bank’s account at the Federal Reserve Bank of New York.
- Gmail, January 2017. A worm-like campaign using a fake Google Docs sharing invitation spread through contact lists and reached a large share of Gmail users before Google shut it down within about an hour.
Two of those three were payment fraud rather than credential theft, which is the same pattern the BEC numbers show today.
How to Stay Safe from Phishing
The old version of this page quantified this with percentages from vendor marketing, including a claim that security software blocks 99.9% of phishing and 100% of malware. Those are gone. The guidance that remains is the part that was never really about statistics.
- Verify unexpected payment requests out of band. Given the BEC figures, a phone call to a known number before moving money is the single highest-value habit here.
- Treat urgency as a warning sign. Pressure to act immediately is a technique, not a coincidence.
- Check where a link actually goes, and be wary of QR codes in email, which exist partly to move you onto a less protected device.
- Use phishing-resistant multi-factor authentication. Stolen credentials are much less useful on their own.
- Do not rely on spotting bad grammar. That tell is largely gone, which is the practical consequence of the AI research above.
- Report it. Those 191,561 complaints are the reason this data exists at all.
Security tooling and training both help, and neither is a solved problem. If you are assessing the wider risk picture, our cybersecurity statistics and data breach statistics cover the ground around phishing.
Conclusion
Phishing in 2026 is not growing explosively, which is the opposite of what this page used to claim. APWG saw 3.8 million attacks in 2025 against 3.76 million in 2024, and Microsoft’s detections fell between Q1 and Q2.
What changed is the method. QR phishing rose 146% in a quarter and then switched delivery format entirely. Smishing is up 40% quarter on quarter. Telecom went from 5.9% of attacks to 33% in two quarters. AI is making targeted attacks cheaper to produce, with a controlled study showing 4.5 times the click-through rate.
And the money still sits mostly next door. Phishing produced 191,561 complaints and $216 million in reported losses, while business email compromise produced a fraction of the complaints and $3.05 billion. Phishing is usually how attackers get in. It is rarely where they get paid.