Table of Contents
The fraud wall: Why agentic orders get blocked
Sophia Willows
Head of Engineering @ Rye
Aug 27, 2025
Merchants routinely block agentic orders as fraud. Rye’s Universal Checkout API adapts with human-like behavior so agents buy reliably at scale.
Many merchants use systems to stop any purchase that doesn’t appear to be from a real shopper, regardless of the intent. This well-intentioned but overbroad filtering ensnares many agents, causing them to fail and fall back on manual intervention.
Fraud detection is one of the hardest problems in agentic commerce. Rye has a solution.
Many merchants use systems to stop any purchase that doesn’t appear to be from a real shopper, regardless of the intent. This well-intentioned but overbroad filtering ensnares many agents, causing them to fail and fall back on manual intervention.
Fraud detection is one of the hardest problems in agentic commerce. Rye has a solution.
Many merchants use systems to stop any purchase that doesn’t appear to be from a real shopper, regardless of the intent. This well-intentioned but overbroad filtering ensnares many agents, causing them to fail and fall back on manual intervention.
Fraud detection is one of the hardest problems in agentic commerce. Rye has a solution.
Why merchants are scarred by bots
Merchants have spent years suffering from automated abuse. Bots aren’t new, and the damage they cause is real. Some are opportunistic, hustling for profit by scooping up products or data. Others are outright fraud. Together, they translate into chargebacks, unrecovered goods, and higher operational costs.
Common examples include:
Scalping: snapping up limited-edition releases before real customers ever get a chance
Scraping: competitors extracting pricing and inventory data at scale
Credential stuffing: breaking into user accounts by reusing stolen passwords
Card testing: cycling through stolen credit card numbers to find one that works
Loyalty and gift card draining: stealing value directly from legitimate customers
Promo, affiliate, and refund abuse: gaming merchant policies to pocket funds
Analysts estimate the annual toll of bot-driven fraud in e-commerce at more than $180 billion, with no sign of letting up as technological advances have led to an entire Fraud-as-a-Service industry.
How merchants defend against bots
In response, merchants and their vendors have built powerful defenses. Modern anti-fraud platforms combine dozens or even hundreds of data points to decide if a session looks human. They check whether the IP address is close to the shipping region, whether traffic comes from a residential network or a data center, and whether the browser presents like consumer hardware. They analyze how a shopper scrolls, moves their mouse, and fills in forms, looking for the natural patterns of human interaction.
These systems are good at what they do. These tools deserve credit for making e-commerce safer and more trustworthy for consumers while helping merchants maintain the margins required to stay in business. Increasingly, vendors are working to separate “good” automation from “bad,” so that not every bot is detected as fraud. But today the filter is still closer to binary than nuanced. And those legitimate agentic e-commerce orders, representing people who want to pay them good money, are so often rejected.
In response, merchants and their vendors have built powerful defenses. Modern anti-fraud platforms combine dozens or even hundreds of data points to decide if a session looks human. They check whether the IP address is close to the shipping region, whether traffic comes from a residential network or a data center, and whether the browser presents like consumer hardware. They analyze how a shopper scrolls, moves their mouse, and fills in forms, looking for the natural patterns of human interaction.
These systems are good at what they do. These tools deserve credit for making e-commerce safer and more trustworthy for consumers while helping merchants maintain the margins required to stay in business. Increasingly, vendors are working to separate “good” automation from “bad,” so that not every bot is detected as fraud. But today the filter is still closer to binary than nuanced. And those legitimate agentic e-commerce orders, representing people who want to pay them good money, are so often rejected.
In response, merchants and their vendors have built powerful defenses. Modern anti-fraud platforms combine dozens or even hundreds of data points to decide if a session looks human. They check whether the IP address is close to the shipping region, whether traffic comes from a residential network or a data center, and whether the browser presents like consumer hardware. They analyze how a shopper scrolls, moves their mouse, and fills in forms, looking for the natural patterns of human interaction.
These systems are good at what they do. These tools deserve credit for making e-commerce safer and more trustworthy for consumers while helping merchants maintain the margins required to stay in business. Increasingly, vendors are working to separate “good” automation from “bad,” so that not every bot is detected as fraud. But today the filter is still closer to binary than nuanced. And those legitimate agentic e-commerce orders, representing people who want to pay them good money, are so often rejected.
The agent's challenge
For most agents, these defenses are an insurmountable barrier. They operate from the wrong networks, behave mechanically, and lack the consumer-like fingerprints fraud systems expect. The result is canceled orders and broken flows.
Often, when an agent gets blocked, a hidden human finishes the order. That keeps the transaction alive, but it slows things down and drives up costs—a single human-in-the-loop cycle can cost up to $1 per order, or even $3 for onshore teams, which will not be recouped if a buyer decides not to buy or gives up on a slow process. That’s not scalable.
For most agents, these defenses are an insurmountable barrier. They operate from the wrong networks, behave mechanically, and lack the consumer-like fingerprints fraud systems expect. The result is canceled orders and broken flows.
Often, when an agent gets blocked, a hidden human finishes the order. That keeps the transaction alive, but it slows things down and drives up costs—a single human-in-the-loop cycle can cost up to $1 per order, or even $3 for onshore teams, which will not be recouped if a buyer decides not to buy or gives up on a slow process. That’s not scalable.
For most agents, these defenses are an insurmountable barrier. They operate from the wrong networks, behave mechanically, and lack the consumer-like fingerprints fraud systems expect. The result is canceled orders and broken flows.
Often, when an agent gets blocked, a hidden human finishes the order. That keeps the transaction alive, but it slows things down and drives up costs—a single human-in-the-loop cycle can cost up to $1 per order, or even $3 for onshore teams, which will not be recouped if a buyer decides not to buy or gives up on a slow process. That’s not scalable.
Rye's approach to fraud detection
Rye’s Universal Checkout is engineered directly for this problem.
Our Merchant Risk Adapter routes every interaction through residential, geo-proximal IPs selected to match the buyer’s shipping region. We also tune browser launch parameters and interaction profiles to resemble consumer hardware and natural usage. Page timings, scroll behavior, mouse paths, and form-filling cadences all look like what anti-fraud systems expect to see from real customers.
We hope fraud defenses will evolve to understand the nuance between good and bad bots, but agentic commerce can’t wait on other industries to continue making progress. We’re confident we’re doing the right thing: These are real people, buying real products, through their chosen agents. It’s no different from someone asking a friend or colleague to buy something on their behalf from the supermarket.
By building technology that looks like humans to fraud systems, we ensure our agent can successfully place orders on humans’ behalf to everyone’s benefit. Merchants get the sale. Implementers close the recommendation loop. Consumers get what they want with less hassle.
Explore the docs and see how the Universal Checkout API ensures that orders get placed—and delivered.
Rye’s Universal Checkout is engineered directly for this problem.
Our Merchant Risk Adapter routes every interaction through residential, geo-proximal IPs selected to match the buyer’s shipping region. We also tune browser launch parameters and interaction profiles to resemble consumer hardware and natural usage. Page timings, scroll behavior, mouse paths, and form-filling cadences all look like what anti-fraud systems expect to see from real customers.
We hope fraud defenses will evolve to understand the nuance between good and bad bots, but agentic commerce can’t wait on other industries to continue making progress. We’re confident we’re doing the right thing: These are real people, buying real products, through their chosen agents. It’s no different from someone asking a friend or colleague to buy something on their behalf from the supermarket.
By building technology that looks like humans to fraud systems, we ensure our agent can successfully place orders on humans’ behalf to everyone’s benefit. Merchants get the sale. Implementers close the recommendation loop. Consumers get what they want with less hassle.
Explore the docs and see how the Universal Checkout API ensures that orders get placed—and delivered.
Rye’s Universal Checkout is engineered directly for this problem.
Our Merchant Risk Adapter routes every interaction through residential, geo-proximal IPs selected to match the buyer’s shipping region. We also tune browser launch parameters and interaction profiles to resemble consumer hardware and natural usage. Page timings, scroll behavior, mouse paths, and form-filling cadences all look like what anti-fraud systems expect to see from real customers.
We hope fraud defenses will evolve to understand the nuance between good and bad bots, but agentic commerce can’t wait on other industries to continue making progress. We’re confident we’re doing the right thing: These are real people, buying real products, through their chosen agents. It’s no different from someone asking a friend or colleague to buy something on their behalf from the supermarket.
By building technology that looks like humans to fraud systems, we ensure our agent can successfully place orders on humans’ behalf to everyone’s benefit. Merchants get the sale. Implementers close the recommendation loop. Consumers get what they want with less hassle.
Explore the docs and see how the Universal Checkout API ensures that orders get placed—and delivered.
Related articles
Related articles
Monetize
your AI platform
with shopping.
Monetize
your AI platform
with shopping.
Monetize
your AI platform
with shopping.
Product
Resources