Key takeaway: Merchants that block all AI agent traffic by default may be closing off a channel that converts. A default-block policy doesn’t just filter out fraud — it can also turn away business. Merchants should weigh that trade-off explicitly rather than treating all bot traffic as risk.
Spreedly executive vice president of product strategy Adam Hiatt told PYMNTS that sessions starting in large language model chat interfaces are converting at a notably higher rate than sessions starting from traditional search. He pointed to one ticketing client where agent-originated traffic converted roughly two to three times better than search traffic.
“The biggest evolution is that not all the bots are bad now,” Hiatt told PYMNTS.
Hiatt described the conversion gap at that ticketing client in stronger terms than a modest lift. “That’s not two to three percent,” he said. “That’s 100% to 200% greater.” He argued that an AI agent arriving at checkout often represents a purchase decision that is “pretty much ready to go,” rather than a browsing session still open to influence.
The figure comes from a single, unnamed client. Spreedly has not disclosed the sample size, time period or comparison method behind it. Readers should treat it as one case rather than a verified industry benchmark.
The data lands alongside a separate warning about how merchants are currently handling this traffic. A Darwinium survey of 500 fraud, risk and security leaders found that 48% allow agentic traffic by default. Another 31% block it by default, and the rest apply some form of case-by-case review.
Chargebacks911 CEO and founder Monica Eaton separately told PYMNTS that many fraud detection systems still rely on human behavior signals — mouse movement, typing cadence and click timing — to flag risk. Those systems are misclassifying legitimate AI shopping agents as bots. That misclassification pushes up false declines on transactions merchants would otherwise want to accept.
Why it matters: These two data points together suggest a risk for merchant fraud stacks tuned purely for bot detection. Such systems could be declining a growing share of legitimate, high-intent agentic traffic. A static block-all-bots policy won’t surface this problem on its own — it just quietly turns a fraud-prevention setting into a revenue leak.