The Digital Commerce Industry’s first agentic-native closed-loop learning platform that helps retailers, brands and agencies evaluate, enrich, distribute, and monitor their products on Agentic Commerce Engines such as ChatGPT, Google Gemini/AI Mode, Perplexity, Meta.ai, Microsoft Copilot, Grok and Claude.
RALEIGH, NC / ACCESS Newswire / September 16, 2025 / ReFiBuy today announced the general availability of its first product called Commerce Intelligence Engine.
“The ReFiBuy product team has been cooking, building a ground-up solution to the challenges Agentic Commerce presents to retailers, brands and agencies,” said Scot Wingo. “I also want to thank our ten design partners who gave us extremely valuable feedback in the last 90 days as we’ve been putting the finishing touches on the product.”
“ReFiBuy’s Commerce Intelligence Engine has accelerated our understanding of our readiness for Agentic shopping experiences and how GenAI understands our product catalog, “said Josh Krepon, President of US Direct to Consumer and Global Digital, Steve Madden. “The platform’s insights and actionable recommendations enhance our content and visibility for our GenAI digital shelf. The product-level GEO monitoring capabilities will play a key role in futureproofing for this new era of commerce.”
Commerce Intelligence Engine’s Core Six Capabilities
At the heart of the Commerce Intelligence Engine is ReFiBuy’s belief that everything in digital commerce starts and ends with the product catalog.
-
Ingest – The Commerce Intelligence Engine starts with the ingestion of the brand or retailer’s product catalog. Our ingestion capability is built with ultimate flexibility pulling the product catalog from any existing system or using web data extraction.
-
Evaluate – Evaluate reviews both catalog data and PDPs for each SKU, flagging GEO structural elements that cause Agentic Shopping Agent issues. Then Evaluate assesses titles, images, variations, and existing attributes, while also recommending missing attributes LLMs need for additional context. All findings are summarized in a scorecard across Content Quality, GEO Structure, LLM Semantic Quality, and User Experience.
-
Enrich – Based on the customer’s feedback and approval from the Evaluate process, Enrich implements the suggested and approved changes at the product catalog level. The product catalog is now optimized to maximize Agentic Shopping Engine exposure.
-
Sync – At any point in the process, customers can export their enriched product catalog to a variety of file formats, or route it directly into existing systems and workflows for further processing or approvals.
-
Distribute – At the customer’s request, the Distribute capability delivers the now enriched product catalog to up to six different Agentic Shopping Engines including: ChatGPT, Gemini, Perplexity, Meta.ai, Microsoft Copilot, Grok and Claude.
-
Monitor – The Commerce Intelligence Engine’s Monitor capability scans the six Agentic Shopping Engines and score’s a product across three criteria: visibility, data quality and AI Digital Shelf.
Evaluate Capability Details
Once ingested, the system evaluates both the site-structure at the PDP level and the product catalog data level for every SKU in the customer’s catalog.
At the PDP level, Evaluate looks at structural issues and makes recommendations that will optimize the PDP for Agentic Shopping Engine’s web crawlers, agents and agentic browsers to access all the product data, reviews and other information that gives the LLM context on the product. Common issues the system detects are stale/incorrect schema.org data, unintentional bot blockers, and the use of complex JavaScript elements that prevent the agents from ‘seeing’ product content.
At the product catalog level, Evaluate examines every data element of the product catalog (from all sources including PDP, data feed or legacy systems) such as titles, images, all variations and existing attributes. In addition to existing attributes, the Evaluate capability recommends incremental attributes to add to the product catalog to give Agentic Shopping Engines a more complete picture of the product’s capabilities, compatibilities, etc.
Evaluate then takes the recommendations from the PDP structure and the product catalog and produces a detailed per-product scorecard based on four criteria:
-
Product Content Quality – Is the existing PDP-level product catalog data content optimized for GenAI? Typical recommendations include adding more context, improving titles, enhancing existing attributes and adding missing attributes through our enrichment capability.
-
GEO Structure – Many retailer and brand sites have structural issues with bot blockers, bad metadata, incorrect or missing llms.txt and frequently schema.org is missing, outdated or has the wrong data.
-
LLM Semantic Quality – Assesses the semantic coherence and natural language understanding quality of the product descriptions.
-
User Experience – Measures the overall user interface, accessibility and interaction quality of the product page.
Monitor Capability Details
Called GEO by other vendors, instead of looking at high-level ‘sentiment’, the Monitor feature takes a subset sample of the customer’s product catalog, goes to out to seven Agentic Shopping Engines (ChatGPT, Gemini, Perplexity, Microsoft Copilot, Meta.ai, Grok and Claude), where it executes product-level prompts designed to replicate shopper behavior. Monitor then records screen shots of the Agentic Shopping Engines’ detailed responses
Monitor then evaluates each Agentic Shopping Engine’s product-level prompt response across three criteria:
-
Visibility – Initially, due to bot blocking, structural problems and other issues, a large portion of a customer’s products are not even ‘indexed’ or trained on by the Agentic Shopping Engines. This is the pre-requisite for showing up at all.
-
Data Quality – Once a product’s visibility has been verified and the product is being displayed by the Agentic Shopping Engine, Monitor verifies all the product information displayed is correct including: in-stock status, variation details, attributes, ratings are all being pulled in correctly, the correct images are shown, and the right product detail page link is provided to the agent or the human user. Also, all pricing information, any free shipping or other offers are verified to make sure the Agentic Shopping is presenting the shopper with the most compelling and competitive offer.
-
AI Digital Shelf – Finally, Monitor evaluates each product’s ‘AI Digital Shelf’. ReFiBuy believes every retailer and brand’s goal should be to have their product ‘own the product card’ meaning that when the human or agent go to buy the product, they win the top spot, and the zero-sum game of the purchase. The AI Digital Shelf reports to the customer their relative position on the product card as well as total transparency around competitor’s positioning, pricing and how they are representing the same products.
Together the scores and details of these three areas give the customer a first-of-its-kind 360-degree view of exactly what the customer experiences are at the Agentic Commerce Engine level.
First Closed-Loop Learning Software System
Once the customer’s product catalog has been through the Commerce Intelligence Engine for the first time, the Monitor three criteria scoring is evaluated and acted on.
If the customer’s product is not visible, if there are data errors or if it does not ‘own the product card’, the Monitor capability alerts the Evaluate/Enrich agents and the customer with details of the problem(s) detected at the Agentic Shopping Engines. The Evaluate/Enrich agents then work with the customer’s input to recommend solutions and then distributes a repaired product catalog to the appropriate engine with fixes for the issues detected by Monitor. Finally, the Monitor capability evaluates that the fix has resolved the problem and if it hasn’t, the cycle repeats until all products are completely optimized. As the Agentic shopping engines change and various product catalog strategies are implemented by ReFiBuy’s agents, the system learns and adapts without code changes.
Complete Flexibility with Existing Systems and Workflows
Founded by ecommerce veterans, ReFiBuy knows that today’s digital leaders at brands, retailers and agencies have invested hundreds of millions in ecommerce platforms PIMs, ERPs and a variety of other systems. Also, modern digital commerce teams operate using complex well-oiled workflows. Because of these existing systems and workflows, ReFiBuy’s Commerce Intelligence Engine is designed to be completely flexible and integrate seamlessly with the customer’s existing systems and workflows.
For example, for our ‘ingest’ and ‘sync’ capabilities instead of bespoke system-to-system integrations, ReFiBuy has built an Agentic integration capability that automatically creates integrations to existing systems, maps the correct fields, tests, deploys and monitors the integration. Or if the customer wants to get started quickly, the system can jump-start using web data extraction.
About ReFiBuy
Founded by Scot Wingo (Founder of ChannelAdvisor/Rithum, co-host of The Jason & Scot Show), Cameron Bowe (ChannelAdvisor), James Frawley (ChannelAdvisor, MikMak), and Derek Conlin (ChannelAdvisor, Walmart), ReFiBuy is an AI-native company focused on revolutionizing digital commerce with Agentic AI. By helping retailers, brands and agencies navigate Agentic Commerce and optimize how consumers Research, Find, and Buy their products.
More information can be found at: www.refibuy.ai
Media resources are available at: https://www.refibuy.ai/media-resources
Media contact: Rebecca Ross – rebecca@refibuy.ai
Social Links:
-
LinkedIn: https://www.linkedin.com/company/refibuy
SOURCE: ReFiBuy
View the original press release on ACCESS Newswire
