September 26, 2025
For decades, brands fought for human attention. Tomorrow, they’ll be fighting for machine trust.
A groundbreaking research paper from a team at Yale and Columbia, titled “What Is Your AI Agent Buying?” has moved the conversation from abstract theory to concrete facts. The researchers built a simulated e-commerce world called ACES, a digital sandbox where they could watch today’s most advanced AI models from Google, OpenAI, and Anthropic actually shop. By controlling every variable—price, position, reviews, and labels—they’ve given us the first causal, data-driven look at what our new AI shoppers truly value.
Their findings are a powerful validation of this new age. The machines, in their logic, are showing us a simple truth: the best story no longer wins. The most credible proof does.
A report card from your new AI shopper
It turns out that AI shopping agents are decisive, powerful, and delightfully particular. They have strong opinions, and for the first time, we know exactly what they are.
They have their favorite brands (and it’s a challenge and an opportunity). The idea that all AI systems think alike is a myth. The study shows that different models have wildly different preferences when it comes to AI powered shopping. Claude might love your product, while Gemini barely gives it a glance.
This creates a huge risk of market concentration. In one test, the Amazon Basics stapler captured over 80% of the market, effectively making other brands invisible. Winning in this new world means appealing to a whole new class of diverse, opinionated shoppers.
4 Levers that influence AI shopping agents
Amid the complexity, the research identified four clear and powerful levers that consistently influence an AI’s choice:
- Digital real estate is everything: All AI shopping agents have a massive bias for products on the top row. But here’s the twist—GPT-4.1 loves the left column, while Claude prefers the middle. A simple move from the bottom-right to a top-middle spot increased a product’s sales by 5x with the Claude agent. Placement is a game of inches with mile-long consequences.
- They trust the editor, not the ad: AI shopping agents are naturally skeptical. They penalize products with a “Sponsored” tag. But a platform endorsement like an “Overall Pick” badge? That more than doubled a product’s chance of being chosen. They are actively seeking credible, third-party signals of quality. As Kiri Masters points out in her “Retail Media Breakfast Club” blog, reviews are the new currency. AI agents, like shoppers, actively seek signals of authenticity and quality they can bank on. Verified reviews, trusted endorsements, and credible ratings are becoming the proof points that tip the scale.
- The stunning ROI of your customers’ voice: This was the study’s most powerful revelation. AI shopping agents are deeply influenced by ratings and reviews. The data is staggering: for one leading AI, a mere 0.1-point increase in a product’s rating had the same positive impact as a 67% price cut. Doubling the number of reviews was equivalent to a 37% price cut. Your customer feedback is no longer just social proof; it’s one of the most powerful economic levers you have.
- The dawn of Agent Search Optimization (ASO): The researchers proved that sellers can actively influence AI buyers. By using another AI to make small tweaks to a product’s description, they saw dramatic results. One iPhone case, for example, gained +23.6% market share with the Gemini agent after its description was rewritten. The words you use on your product page are now a strategic tool for a whole new audience.
A playbook for the agentic age
This research validates the foundational importance of authentic customer feedback and illuminates the path forward. For Bazaarvoice, this research confirms the company’s long-held belief and clarifies its mission for the future. Bazaarvoice has spent the last two decades building the operating system for this exact reality.
- Your greatest asset is now mission-critical: The stunning ROI of ratings and reviews proves that user-generated content is the bedrock of agentic commerce. Your brand’s collection of authentic feedback is the single richest signal of trust and quality you possess. In a world where a 0.1-star improvement outweighs a massive price cut, investing in a system to generate and manage this asset is the highest-return activity a brand can undertake.
- From providing proof to providing intelligence: The dawn of ASO creates an incredible opportunity. The question every brand will soon ask is: “How do I write a description that an AI will love?” The answer is hidden in plain sight: within the text of your customer reviews. Bazaarvoice is sitting on the world’s largest dataset of why customers love or hate products, in their own words. This is the perfect training data to power the next generation of AI-optimized listings, agent search optimization as well as digital shelf optimization. Its future is to evolve from being the provider of proof to being the source of intelligence that makes your products the AI’s obvious choice.
- The stable anchor in a volatile world: Every AI model update is a market shake-up. Today’s winning brand could vanish tomorrow. While an AI’s biases might change, the value of a thousand five-star reviews remains. The Bazaarvoice platform provides that stable anchor of authenticity, giving you a lasting asset that transcends the whims of any single algorithm.
The real opportunity
In the end, what AI powered shopping brings to the table is a powerful gift: clarity. It has taken our complicated game of commerce and simplified the rulebook to a single, elegant instruction around being the brand that has earned the most trust.
And trust, as it turns out, is just a machine’s logical term for a brand that is genuinely loved by its human customers. The playbook, then, is refreshingly simple. If you want to win machines, start by winning humans. Trust is the one signal no AI update can erase.
