ChatGPT did not provide any product links in its initial responses. But he easily provided them when I asked, and although I didn’t click on every one of them, none of them seemed like hallucinations. Claude, on the other hand, apologized and stated that it “cannot actually contain direct links to websites or products.” Anthropic hasn’t yet rolled out a web search feature for Claude, but the company says it’s working on it.
This technically made Claude the least useful chatbot I tested for shopping. But it also means that Anthropic has so far avoided treading into the ethically murky territory of allowing its AI-powered chatbots to pull human product reviews from the web. Instead, Claude bases his product comparisons on an existing data set. Embarrassment, meanwhile, claims that with Buy by Pro, people “no longer have to sift through countless product reviews.”
When I asked Perplexity what I should buy for my editor/musician friend, he recommended a set of solar-powered bike lights (I also noticed he was a cyclist). It wasn’t a bad idea, but it wasn’t a milestone-worthy birthday gift. I kept tweaking my prompt. How about a personalized leather guitar strap? I went down the rabbit hole.
I was beginning to understand that Perplexity’s goal in improving shopping features wasn’t just about helping me brainstorm modern ideas or come up with uniquely thoughtful gifts. Perplexity is playing a longer role, slowly drawing our attention away from competitive corners of the web, gaining a better understanding of how people like me apply its platform and feeding that data into ever-evolving artificial intelligence models. Whenever I needed to refine my search because the initial results were often unsatisfactory, I stayed in the Perplexity app, which meant I was neither on Amazon nor Google (though I eventually ended up on both sites). Perplexity Pro is not a full-fledged e-commerce site, nor is it yet “agent” in any way, but I am one of millions of people providing the information needed to become that thing.
When I turned to Google Gemini, it turned out that the gifts suggested for my 16-year-old niece weren’t bad in themselves, they were just uncreative and, in one case, misleading. It said I should buy her a “blanket to snuggle up with a good book,” but it wasn’t clear whether the blanket was for her or her cat. Kindle was a good idea. But I’m afraid of what she would write to me if I sent her the SAT study guide that Gemini suggested (probably a “thx” and nothing else). My editor/musician friend’s app ideas were equally uninspiring, including “Vinyl Records” and “High Quality Headphones.”
I’ve been using the one-year version of Gemini, but earlier this month Google started rolling out a newer version, Gemini 2.0, to developers and constrained testers. The modern artificial intelligence model “will think many steps ahead and take actions on your behalf,” the company says says. For now, it means taking action on behalf of developers – taking the next step in their coding processes – but I look forward to the day when it can be on my shopping list.
Eventually, ChatGPT led me to an online spice store where I purchased some specialty baking ingredients for my friend, who at that point I thought would be a finalist The great British bake. After all, I’ve been talking to the AI bots for so long that many of the gifts I’ve selected won’t arrive until after Christmas. My niece will receive cash on the card. My search for a birthday gift for a friend yielded no results. I decided to postpone this task until January, a month full of modern things and the agent’s determination.