Wilgot raises $3 million to help retailers optimize for AI shopping
The Stockholm startup aims to turn customer reviews and queries into structured data that helps AI assistants recommend products more accurately.
The short version
- Wilgot raised $3 million (SEK 30 million) in a seed round led by Industrifonden.
- The platform uses 'Agentic Product Optimization' to turn reviews and support tickets into product data.
- The system uses reinforcement learning to adjust content based on visibility and conversion signals.
- Early results show a 20% net sales increase for one customer and 7% of traffic from ChatGPT for another.
- The startup supports content generation in more than 100 languages.
Why it matters
As shopping moves toward conversational AI, brands need a way to ensure their products are accurately represented in AI responses. Wilgot provides a method for retailers to bridge the gap between raw customer feedback and the structured data AI systems require.
Bridging the gap to AI shopping
Retailers often struggle to appear in the results of conversational AI assistants because standard marketing descriptions lack specific context. When a shopper asks an AI for a gift for a hiker, the system needs to know which specific items fit that activity and why. A simple photo or a brief blurb often fails to provide the depth required for an AI to make a confident recommendation.
Wilgot addresses this by focusing on what it calls Agentic Product Optimization (APO). This software is designed to help retailers make their products easier for AI shopping assistants to find and understand. Instead of just creating longer descriptions, the platform extracts relevant facts and attaches them to the correct items to provide more useful evidence for search systems.
The company recently raised $3 million in a seed funding round to expand its reach and develop its platform. This capital will allow the Stockholm-based startup to scale its commercial efforts and refine its technology. The round was led by Industrifonden and included participation from several other venture firms and angel investors.
Turning feedback into data
To build these richer product profiles, Wilgot analyzes information that retailers already possess but may not be using effectively. The platform looks at search queries, customer reviews, support tickets, and return records to identify purchase motivations. These sources can reveal recurring questions or mismatches between what a customer expects and what a product actually delivers.
For example, if many customers ask about the specific fit of a shoe, Wilgot identifies this as a gap in the existing sizing information. The platform then turns these findings into structured data, such as product pages, feeds, FAQs, and specific attributes. By making verified guidance explicit, the company helps both the shopper and the search system access more accurate information.
The goal is to create a system that distinguishes between a verified specification, a customer's personal opinion, and an inference about suitability. This distinction is critical because making an item more visible without accurate data could cause unsupported claims to spread further. Wilgot aims to provide a layer of optimization that changes the underlying information these assistants work with.
A reinforcement learning loop
Wilgot's platform operates through a three-stage process involving demand signal discovery, content generation, and performance measurement. The system generates titles, descriptions, and attributes in over 100 different languages. It then monitors how this content performs across Google and various AI search platforms to see if it improves visibility and conversion.
The company uses reinforcement learning to improve product information based on this performance feedback. This means the system can learn which specific content changes are actually useful after they are published. While the specific reward designs and model architectures are not public, the loop is designed to handle massive catalogs containing tens of thousands of products.
It is important to note that this process does not allow a retailer to directly retrain a model like ChatGPT. Instead, it optimizes the records that these independent assistants use to generate their answers. By improving the quality of the source data, retailers can influence how they are represented in the broader AI ecosystem.
Early results and team
The startup has already begun working with several notable customers, including Daniel Wellington, J.Lindeberg, Myrqvist, and Signify. Early results from these partnerships show varying levels of success. For instance, net sales increased by 20% across Daniel Wellington product pages optimized by the platform.
Another customer, Myrqvist, reported that 7% of its customers now arrive through ChatGPT. While these figures provide early commercial signals, the company has not yet disclosed the specific measurement periods or experimental designs. These results suggest that the platform can successfully drive both sales growth and new customer acquisition channels.
The founding team brings diverse expertise to the project, including a background in deep reinforcement learning and data science at companies like Klarna and IKEA. CEO Erik Wikander noted that brands currently spend too much time measuring AI mentions and not enough time giving AI a reason to recommend their products. The new funding will help them scale this approach to more retailers.
What to watch
- How other retailers will adopt APO to compete for AI search visibility.
- Whether the 7% of ChatGPT traffic for Myrqvist scales as more brands join the platform.
Try it
Look at your own recent online purchases and see if the product description actually answered the specific questions you had before buying.
Sources and links
- ReportingUnite.AI: the article we sourced
First reported by Unite.AI. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.