I Tried Meta AI's Shopping Assistant, and I Won't Be Using It Again

I tend to go out of my way to avoid using Meta AI, but today, I gave it a fair shake. That's because, as reported by Bloomberg, Meta's AI service is now testing a shopping assistant. The idea is to compete with similar services from AI platforms like ChatGPT or Gemini, where users tap into the power of generative AI to search the web for product recommendations. That's all fine and well in theory, but in execution, Meta AI's shopping assistant is a bit of a disaster—even if it is only in testing.

Meta AI's shopping suggestions are useless

When I tried accessing Meta AI's web app logged out, I didn't see the shopping assistant. But once I signed into my Meta Account, a number of additional options appeared—including "Shopping." When you click the option, a new "Shopping research" option appears in the model selector, alongside "Thinking" and "Fast." While you can type your own prompt into the search bar, a number of suggestions appear below, too. At this time, I see things like "Help me find a signature scent," "Show me mid-century modern living room furniture," or "Find a one-of-a-kind gift for my friend." Finally, AI is here to help you spend money when you can't think of anything to buy.

I was pretty intrigued by that last prompt. Meta AI doesn't know my friend, and yet, it's going to find me a "one-of-a-kind" gift they'll love? I had to see what the bot would come up with. After Meta AI thought for a while (11.6 seconds, or so it tells me), the bot indeed confirmed my suspicions, asking for more info about my friend. It needed details like hobbies and interests, budget, and special memories I could incorporate into the gift. You would think that some gift ideas of my own would pop up by just thinking through these questions myself, but I needed the AI's help, so I offered the following made-up answers to test it out: "Their hobbies and interests are music, biking, and going to the movies. My budget is $100. We went on vacation to Hawaii and we missed our flight home."

The AI thought for 45.8 seconds, before returning with its "one-of-a-kind" gift ideas. Its favorite was something that combined my "friend's" interest in music and our fake vacation to Hawaii: a ukulele. One-of-a-kind, indeed. Meta AI reasoned that, "a Ukulele Starter Kit is the perfect way to bring the island vibes home. The Kala Learn to Play Ukulele Concert Starter Kit is right at your budget and comes with everything they'd need to start, including lessons and a tuner. It's a great hobby for a music lover and a direct connection to Hawaii." Other unique gifts included three different types of portable speakers. Thanks.

"For the movie buff," Meta AI suggested a scratch-off poster of 100 movies everyone should see. Or, as a "subtle nod to our trip," a movie poster of film shot in Hawaii. This, I'll admit, is a bit more unique, even though I'd never actually buy that. After scrolling past some bike accessories, though, I really hit gold:

"A Custom Hawaiian Shirt would be hilarious," said the generative AI chatbot. Credit: Lifehacker

Meta AI isn't much more helpful if you search for products yourself

After failing to find a perfect gift for my fake friend, I gave a custom prompt a try next: "I need a new couch for my living room." This was a bit more helpful. The bot returned a list of five different couches, each with a description of the design and brand, in addition to a carousel of 12 couches of varying prices. To narrow it down, the bot asked me some questions about my home and interests, like how big my living room was and what style I was looking for. I sent the bot the following: "My living room is 200 sq. feet. I want a modern leather couch under $2000." This returned what seemed to be a relevant list of couches, each within budget and style. I also learned about some new brands I didn't know before the search, like Article and Poly & Bark.

Finally, I asked Meta AI for help finding a new MacBook. I said I needed something fast, but under $1,200. It thought, then brought back three decent options: a 13-inch M4 MacBook Air for $999, a 15-inch M4 Air for $1,199, or an M3 MacBook Air with extra storage for $1,030. There's nothing wrong with those suggestions, but there were two issues I found with the result. First, the link for that third MacBook Air recommendation didn't actually go to a listing, but the homepage for Abed Tahan, a store based in Lebanon. A quick search of the site returned results for the M3 MacBook Air in question, but it was more expensive than Meta AI said it was, and the store doesn't ship to the U.S. Second, I asked Meta AI whether these results were the latest newest models, and it confirmed they were—despite the fact that Apple had announced new M5 MacBooks this morning. If the bot was working on a limited knowledge base, that'd be one thing, but it performs web searches with each query. It should be able to find this information.

While there are some apparent plusses to Meta AI's shopping assistant, like its ability to find furniture that fits a particular room size and style, I feel pretty confident that I won't be using it. The fact that some products links don't actually work, and that it can't reliably give you the most up-to-date products on the market, defeats the entire purpose of a shopping tool. I'll be sticking to my usual shopping research: a traditional search engine combined with real user experiences.

View the full article