At A Glance
Today’s AI-powered fraud bots are exceptionally good at mimicking human behavior and slipping past the device and network checks that would’ve revealed earlier generations. Many record and replay real users’ mouse movements and keystrokes, rotate through thousands of IP addresses and adapt when they hit a control. For businesses that see high application volume every day, that’s a dangerous combination: a wave of fraudulent applicants can enter the ecosystem without causing a noticeable ripple in total volume.Fraudsters have put generative AI (GenAI) to work in more ways than one. AI-generated deepfakes and fake documents get the most attention, and for good reason: they’re eerily convincing, built to fool both people and the checks designed to verify them. Experian’s2026 Future of Fraud Forecasteven warns of deepfake job candidates capable of passing interviews in real time.
But beneath the surface, AI is changing something more fundamental: how fraudsters launch attacks. Work that once took skilled teams and weeks of preparation, such as writing attack scripts, probing a target’s defenses and adjusting when a control blocks them, can now be automated and refined in minutes. That gives fraudsters the best of both worlds: the scale of a brute-force ambush with the sophistication of their most advanced, bespoke attacks.
Much of that shift runs through fraud bots. At the start of 2024, malicious bots made up 30% of the fraud attacks observed by Experian. By the end of the year, they made up 80%. The trend has only accelerated since: AI-enabled bot attacks grew 12.5x year over year in 2025, according to the2026 Thales Bad Bot Report.
One attack shows how easily these fraudsters slip through the cracks, especially at fast-growing digital businesses.
Fraudsters hiding in the shadows: A real-world example
Merchant onboarding companies face fraud every day. Many prioritize easy onboarding over fraud prevention; some expect up to 15% of applications to be fraudulent as a baseline, and a single attack can mean millions in losses.
One merchant-onboarding fintech had grown accustomed to massive application volume and, with it, a steady stream of fraudulent applicants. Fraud typically held steady, while overall volume followed a peak-and-valley pattern as businesses applied during working hours.
Then came a week of record-high applications.On the surface, it was business as usual. Total volume followed its familiar pattern, in line with day-to-day trends. If the fintech had only watched for abnormalities in total volume, nothing would have looked out of place. In fact, the surge would have read as a welcome sign of growth, maybe even a reason to relax defenses further.
ButNeuroID,Experian’s behavioral analytics solution, revealed afraud ringhiding under the wave of applicants. With crowd-level visibility into user behavior, the fintech spotted a spike in its normally consistent fraud volume during the surge. It wasn’t dramatic- a jump of 13 percentage points in the share of risky applicants- but left undetected; the coordinated attack could have cost millions. Using NeuroID, the fintech found the gaps in its onboarding flow that fraudsters were exploiting and closed them without adding friction for genuine customers.
Not every bot is a bad bot
Spotting fraud botshidden in normal traffic is hard enough. But the emergence of consumer-ready AI agents presents a second challenge: not all automated traffic is fraud.
Consumer artificial intelligence (AI) platforms such as ChatGPT, Gemini and Claude now offer agents that browse websites, fill out forms and complete transactions on a real person’s behalf. Customers are using them to research products, shop, open accounts, and manage their finances. Those agents run on many of the same tools as fraud bots: automated browsers, machine-speed data entry, and none of the hesitation that marks a human session. And fraudsters can use the same consumer tools, or spoof a trusted agent’s identity, to blend in.
That changes the job.Blocking every botused to be the safe answer, because bots rarely had a legitimate reason to be on a sign-up or login page. Today, a blanket block can turn away customers and the revenue that comes with them. And as agent traffic grows, it becomes another surge for fraud bots to hide in, just like the record week at our merchant-onboarding fintech.
The question fraud teams need to answer has shifted from “Is this a bot?” to “Is this bot acting for a real customer, or committing fraud?” Most bot detection tools weren’t built to make that distinction.
Modern fraud bots need modern defenses
GenAI fraud is coming from every angle, and malicious tools like FraudGPT arewidely available. If you’re only stopping bots by looking for lightning-fast data entry or telltale device and network red flags, it’s time to rethink your approach. Even if your fraud stack is holding its own today, fraudsters will evolve to beat it.
Experian’s NeuroID helps you separate humans from bots, and good bots from bad, in real time:
And when agents are ready to transact,Experian Agent Trustverifies the person behind the payment, the legitimacy of the agent and the authorization of the payment.
Gain more insight into NeuroID’s ability to uncover large-scale attacks in real time and how it can assist with AI fraud detection to protect your business from fraud bots without shutting out the customers who send agents on their behalf.