DeepLumen reaches 1,000 merchants as it rolls out pay-on-sales AI commerce
Joy Wu's Shopify app creates machine-readable product pages, tracks AI-attributed orders and charges merchants when those sales happen.
By RuntimeWire Staff · Published
Primary source: PR Newswire
Why it matters
DeepLumen is testing whether AI visibility software can become a transaction business. A working cost-per-sale model would let Wu acquire merchants with a free app and monetize verified sales.
DeepLumen co-founder and COO Joy Wu is turning her machine-readable storefront thesis into a performance business. DeepLumen said on August 13 that Agentic Page is live across more than 1,000 merchants, has begun attributing orders from AI discovery and is entering commercial rollout with a cost-per-sale model.
Wu has framed the product around a straightforward prediction: brands will need one storefront for people and another for machines. In an essay, she wrote that "every brand will need two storefronts," with product details, policies, trust signals and transaction rules presented in a form that AI systems can reliably interpret.
Agentic Page is DeepLumen's first product built around that idea to reach commercial scale. DeepLumen generates structured versions of a merchant's product pages so AI assistants can index and recommend items without relying on the visual layout built for human shoppers. The existing storefront remains in place, while the additional pages act as what DeepLumen describes as a machine-readable mirror.
A free app becomes a revenue funnel
DeepLumen publicly introduced Agentic Page on February 13. Its Shopify App Store listing records a May 6 launch and currently lists the product as free. The move to cost-per-sale pricing gives DeepLumen a way to keep installation friction low while collecting revenue when it can connect an AI referral with a completed purchase.
DeepLumen says its fastest merchants received a first attributable AI-driven order within two days of deployment. It also reports AI crawler traffic increases of as much as 659% in selected cases. Both figures are best-case results supplied by DeepLumen. The announcement does not give an aggregate order count, gross merchandise value, median time to first order, merchant retention rate or the percentage DeepLumen will charge under its cost-per-sale model.
The 1,000-merchant total measures distribution. DeepLumen's homepage separately advertises more than 500 Shopify stores, 750,000 AI referrals tracked and an average 4.2 times increase in AI traffic. The 500-store and 1,000-merchant figures may refer to different groups or reporting dates, although DeepLumen does not define the difference. That makes attributable sales the more consequential metric as Wu moves the product from free distribution toward revenue.
Performance pricing raises the stakes for attribution. A merchant will need confidence that DeepLumen caused or materially assisted a sale before accepting a fee on the transaction. AI shopping journeys can cross an assistant, a browser, search results and direct visits, creating room for competing claims over which channel deserves credit. DeepLumen's economics depend on making that chain legible to merchants.
Wu is starting where outcomes can be counted
Wu chose e-commerce because product discovery can lead to a checkout event that software can record. Agentic Page gives AI systems structured catalog information, including attributes that may be scattered across descriptions, shipping pages and return policies. DeepLumen says installation takes about 15 minutes and preserves the merchant's canonical tags so the new pages do not replace the primary website in search indexing.
The current product is built around Shopify, with WooCommerce support planned next. DeepLumen is also developing Agentic Sales, an AI shopping assistant for intent recognition and cross-selling, and the Open Commerce Protocol for agent-mediated transactions. DeepLumen says merchants remain the merchant of record when its protocol is used and that it does not handle their funds.
That product sequence follows Wu's thesis from discovery to purchase. Agentic Page makes products readable. Agentic Sales tries to match shopper intent with inventory. The protocol layer is intended to let an agent complete the commercial action. DeepLumen still has to show that a meaningful share of AI visibility turns into transactions, but charging on sales gives Wu a direct test of the idea.
Agentic commerce infrastructure is filling up
DeepLumen is entering a market where startups and established payment providers are competing to control the data between a merchant's catalog and an AI shopping agent.
ReFiBuy raised $13.6M in May for software that evaluates, enriches and distributes product data at the SKU level. Channel3 raised $6M in December 2025 to build a product graph for developers and merchants. Limy raised $10M in January for AI visibility, conversion and measurement software aimed at larger brands.
Payments providers are moving closer to the same problem. Adyen introduced Adyen Agentic in June, offering product-feed, cart and payment APIs intended to connect merchants with multiple conversational commerce platforms.
DeepLumen's route is centered on a free Shopify installation followed by outcome-based fees. That can give Wu faster merchant distribution than an enterprise sales process, while putting the burden of proof on DeepLumen's attribution system. The 1,000 reported deployments establish reach. The durability of Wu's model will come down to how many of those merchants receive repeat AI-driven orders and accept DeepLumen's calculation of the sale.