Spreedly Sees Good Bots Becoming an Agentic Commerce Channel

Watch more: Digital Shift With Spreedly’s Adam Hiatt

Bots are bad. That, at least, has been digital commerce’s operating assumption since the category’s inception. Automated activity was associated with credential stuffing, scraping, inventory hoarding and fraud. Legitimate customers, it went without saying, were assumed to be human ones. As a result, merchants have traditionally treated automated traffic as something to filter, challenge or block. But as artificial intelligence (AI) moves from recommending products to initiating transactions, agentic commerce is beginning to change that assumption.

“The biggest evolution is that not all the bots are bad now,” Adam Hiatt, executive vice president of product strategy at Spreedly, told PYMNTS.

The commercial challenge, Hiatt said, is no longer simply identifying whether a visitor is a bot. It is determining whether that bot represents a legitimate customer with genuine purchase intent and then making it easy for that agent to transact.

That distinction sounds simple. Its implications for payments, fraud and customer acquisition are not.

Agentic Offers Merchants a New Bottom-of-the-Funnel Traffic Source

The near-term agentic opportunity is emerging inside large language model (LLM) platforms, where consumers can ask an agent to find a flight, product or service within specified parameters and eventually execute the transaction.

“Merchants will need to make their catalogs both available to those AI platforms, and crucially, the payments rails available to those platforms,” Hiatt said, emphasizing that agentic payments may be less about adding another checkout option than opening another distribution channel.

Early evidence suggests AI-originated traffic may behave differently from conventional search traffic, largely because AI agents potentially compress a part of the purchasing funnel that merchants historically controlled themselves.

Hiatt said one of Spreedly’s own ticketing customers has seen sessions originating inside LLM model chat interfaces convert at roughly two to three times the rate of sessions beginning with traditional search.

“That’s not two to three percent,” he said. “That’s100% to 200% greater.”

A consumer arriving through search may still be researching, comparing alternatives or reconsidering whether to buy. An AI assistant, by contrast, can evaluate price, availability, preferences and other criteria before ever reaching the merchant.

“When the bot hits the site, the decision’s pretty much ready to go,” said Hiatt. “It changes the nature of conversion.”

Payments Infrastructure Moves Into the Discovery Layer

If agentic commerce’s early conversion patterns persist, agentic traffic could develop into a distinct acquisition channel alongside search, marketplaces, social media and direct traffic. An agent arriving at checkout may resemble a highly qualified lead more than a traditional website visitor. The merchant’s job becomes less about persuading that shopper and more about ensuring the transaction can be completed.

“Your conversion questions are around, is your catalog right and accurate, and do you have the best representation within these environments?” Hiatt said, adding that merchants accustomed to optimizing landing pages and checkout funnels may have to optimize the information AI systems use before a transaction ever reaches their websites.

Catalog accuracy, product availability and machine-readable information become commercial infrastructure. Search-engine optimization does not disappear, but it gains a new counterpart: ensuring a merchant is sufficiently visible inside the models consumers ask to make recommendations.

Brand, Hiatt said, could remain surprisingly important. Consumers may delegate discovery to AI without surrendering final approval, particularly for expensive or consequential purchases. In that environment, familiar merchant brands can become a trust signal when the underlying artificial intelligence platform itself is not fully trusted.

The Good Bot is Becoming Fraud’s New Problem

Payments providers, orchestration platforms and fraud vendors are racing to build the infrastructure connecting merchant catalogs, payment credentials, agent identities and risk decisions. That means many merchants will not build agentic infrastructure themselves; their competitive decision will be which partners can connect them to emerging channels quickly and safely.

Hiatt described the coming change as a “tidal wave,” with standards and infrastructure acting as the remaining constraints on adoption.

“The ability to create nuanced risk policies that aren’t simply one size fits all” is becoming important, he said, adding that “that fraud orchestration problem is deeply crucial.”

That question turns fraud infrastructure into revenue infrastructure. The merchant that cannot distinguish a legitimate purchasing agent from abusive automation may have only two unattractive choices: allow too much traffic and increase fraud exposure, or block too aggressively and suppress sales.

Watch the full PYMNTS TV episode with Spreedly’s Adam Hiatt to hear more about:

• Why bot traffic is becoming a revenue channel, not just a fraud signal. Hiatt says one ticketing customer is seeing sessions that begin in LLM chat interfaces convert at roughly two to three times the rate of sessions originating in search, suggesting agent-driven traffic may arrive with unusually high purchase intent.
• Why merchants will need to distinguish “good bots” from bad ones. As legitimate AI agents begin initiating purchases, blunt bot-blocking strategies risk suppressing revenue. Hiatt argues merchants will need more nuanced fraud orchestration and emerging “know your agent” capabilities to decide which automated traffic should be allowed to transact.
• Why conversion is moving upstream into the AI interface. When agents compare options and make decisions before reaching a merchant, winning the sale increasingly depends on whether a brand, catalog and payment rails are visible and executable inside AI platforms—turning LLM discoverability into a new part of the commerce funnel.

The post Spreedly Sees Good Bots Becoming an Agentic Commerce Channel appeared first on PYMNTS.com.