AI shopping lessons from an Amazon Rufus veteran

Happy Wednesday.

Yesterday, a federal appeals court handed Perplexity a win in its fight with Amazon over the Comet AI browser that Amazon initiated when it sued the AI startup back in November.

I've been following this case since then because of what it could mean for AI assistants and agents and how they’ll be allowed to access websites and apps on our behalf.

The Ninth Circuit struck down a lower court injunction that had aimed to block Perplexity's Comet AI shopping assistant from accessing Amazon.com. The core issue was whether Perplexity itself "accessed" Amazon and violated the Computer Fraud and Abuse Act, as Amazon has claimed in its suit. The judges said no.

They found that the AI assistant takes a screenshot of a human user's browser, sends it to Perplexity's servers, and gets instructions back. Perplexity's computer systems never actually touch Amazon's servers directly.

Perplexity's chief communications officer, Jesse Dwyer, told me the company would keep fighting for "the right of internet users to choose whatever AI they want." Amazon spokesperson Lara Hendrickson said Amazon disagrees with the ruling and remains confident in its case while evaluating next steps.

Amazon also struggled to show real harm from Comet's actions, with the appeals court calling their evidence “weak.” Amazon can still try to block AI bots through its own terms of service, the court said. The case now heads back to the district court.

The judges were explicit that this ruling only covers Perplexity's Comet assistant in this instance, not agentic AI technologies more broadly. Still, it's worth paying attention to because it's one of the first rulings from a federal court on how an AI assistant can or can't access another company's website. This is the beginning of a legal conversation that will likely last years.

Elsewhere, ChatGPT’s advertising business expansion now appears to include product carousels. Who’s trying this? Let me know.

Speaking of advertisers, the IAB this week published a new report titled “Measuring Visibility in the AI Era.” It’s worth a read if you’re working at a retailer, brand, publisher, or a company offering AEO or GEO services.

Now on to the good stuff…

I don't remember exactly when I first spoke to Mukesh Jain, but I could quickly tell he understood both the technological and consumer behavior challenges of e-commerce.

That makes sense considering Jain spent nearly five years at Walmart, where he was a product lead when the retailer launched its online grocery business, then led product teams overseeing search and personalization. He spent the next eight years at Amazon, where he ran a science, product, and engineering team that built Amazon's "Product Graph," and later held a similar leadership role on Amazon Rufus, the AI shopping assistant recently renamed Alexa for Shopping.

Jain left Amazon last year and is now building SignalAI, a startup that helps job seekers identify and close AI-related skill gaps, and provides a credential proving they're capable of what they say—starting with those who want to be AI engineers. But he’s also a subscriber of The Aisle, and he wrote to me last week with a lot of thoughts about my piece on the AI startup Daydream's move to power conversational search on other brands' websites, not just its own site and app. (In short, he likes the move for them.)

His private takeaways reminded me of how much knowledge he has about where product search has been and where it's going. So I asked him if he'd chat more with me, in hopes that all of you building at the intersection of AI and commerce might get some value from his experiences and insights too. Luckily for all of us, he obliged. The conversation has been edited and condensed in some spots. Most of the discussion happened over video chat, but Jain did follow up with some additions and clarifications by email.