Shopping is beginning to move away from retailer websites and into conversations with AI. Trustpilot believes the change increases the value of its review data, because an agent choosing on a consumer’s behalf needs evidence about which businesses deserve trust.
Chief executive Adrian Blair says the company is pursuing partnerships with major ecommerce platforms and model providers. Referral patterns support the opportunity: click-throughs from AI-based search rose 1,490% over the past year, and Promptwatch ranked Trustpilot as the fifth most cited domain in ChatGPT in January.
The shift creates a new commercial channel for reviews—but also changes who controls the customer relationship.
Reviews become infrastructure for shopping agents
A traditional shopper moves between search results, product pages and review sites. An AI agent can combine those stages, comparing products, assessing sellers and completing checkout inside one interface.
That convenience depends on reliable external signals. A model may know a product’s specifications, but it also needs current evidence about delivery, service and customer experience. Trustpilot’s user-generated reviews can supply context that merchants’ own descriptions do not.
Blair argues that reviews remain valuable regardless of whether a person or agent reads them. Consumers still have experiences with businesses, and the accumulated record becomes an input to future purchase decisions.
The company expects its operating margin to reach 30% by 2030, with LLM use of its content contributing to improvement. Increased citations and referral traffic suggest AI systems can become distribution partners rather than merely alternatives to conventional search.
The commercial challenge is turning visibility into durable revenue while ensuring models use the data accurately and with enough context.
Checkout moves inside the conversation
Major platforms are already building the surrounding infrastructure. Walmart’s relationship with Google allows purchases inside Gemini. Google has similar arrangements with Shopify and other retailers, while Shopify’s Universal Commerce Protocol lets agents access product information and take transactions through checkout.
Microsoft and PayPal are pursuing Copilot Checkout, and Shopify describes “agentic storefronts” where the shopping journey remains inside the AI interaction. Amazon and OpenAI have also announced work involving customised generative systems on AWS for Amazon’s consumer applications.
For shoppers, the appeal is fewer steps. For merchants, the channel may generate sales without delivering the same customer data as a visit to the retailer’s own site. Marketing teams must weigh incremental revenue against weaker ownership of browsing behaviour and the relationship after purchase.
Amazon illustrates the strategic tension. It is reportedly resisting unauthorised third-party agents while developing its own assistant, protecting user data and advertising income rather than allowing another platform to mediate access freely.
Trust becomes both product and bargaining power
As AI platforms become the place where research begins and transactions end, independent information providers gain leverage. A shopping agent that lacks trustworthy business data risks recommending poor sellers and damaging confidence in the entire experience.
Trustpilot’s position is not guaranteed. Review datasets need protection against manipulation, clear attribution and mechanisms for representing disagreement. An agent may compress hundreds of nuanced reviews into one recommendation, making the quality of synthesis as important as the quantity of source material.
Retailers also face a strategic choice. Joining agent platforms can make products discoverable in a growing channel. Refusing access preserves control but may make the merchant invisible when consumers no longer begin with a conventional search.
Trustpilot’s rising AI referrals show that the transition is already changing traffic. Its opportunity is to become a trusted data layer across model vendors rather than a destination users must visit directly.
The wider ecommerce lesson is that the interface is separating from the transaction. AI may own the conversation, a commerce protocol may handle checkout, a retailer fulfils the order and a review platform supplies trust.
When those roles split, data providers can become as important as storefronts. The decisive question is who captures value from the recommendation—and who the consumer holds responsible when the agent chooses badly.
