The Seattle-based chain is testing a beta app in ChatGPT that helps users discover drinks based on vibes and gut feelings alone, the company announced Wednesday, marking another step in corporate America’s bid to make business-facing activities feel less like a menu search for customers.

The beta app is built directly into ChatGPT, where users can enter a prompt describing their caffeine-laced needs and wants. Asking for “something sweet and nutty” can generate a pistachio latte recommendation, for instance. Users can also upload photos and ask the app which drinks would aesthetically complement their clothes or environment.

Starbucks joins retail and travel booking services that have already rolled out similar features offering personalized recommendations to users. Companies are looking to slowly chip away at the burden of choice that comes with shopping, travel, and dining, but in relieving customers of their decision-making, firms might run into some unexpected risks.

Take Walmart as an example. Last year, the company announced a partnership with OpenAI that would allow some customers to shop for Walmart products and check out directly with ChatGPT. The program was first available for around 200,000 Walmart products, and executives were looking most closely at conversion rates, the percentage of shoppers that end up actually making a purchase. 

The trial was a flop. Conversion for products suggested by ChatGPT were three times as low for items that had to be bought through the app than for products that required clicking through to Walmart’s main site, according to a Wired report last month. 

The company has pivoted to embedding its own AI chatbot into ChatGPT to streamline procedures, but shoppers still seem to value the decision-making that accompanies shopping. Daniel Danker, an executive overseeing AI products at Walmart, told Wired that one of the experiment’s shortcomings was that shoppers tend to decide on buying accessories alongside their main purchases.

Discovery or repetition

Starbucks’ app will be different in that users can’t make a purchase directly through ChatGPT, but it could still change the way consumers find their next favorite item, perhaps for the worse. 

Some research has suggested AI recommendations might even be taking some of the excitement out of shopping by browsing. One 2025 study from a university in China, published in the journal Advances in Consumer Research, found that AI-generated suggestions on e-commerce platforms actually tended to lead to more negative feelings among shoppers. While some users appreciated greater simplicity, many reported feeling stuck in “information cocoons,” where AI assistants provided repetitive recommendations that reinforced existing preferences, rather than prompting them toward interesting or unexpected products.

Starbucks representatives say its app can avoid this pitfall by reacting to feelings described or environmental cues rather than user history. “We want to meet customers right in that moment of inspiration and make it easier than ever to find a drink that fits,” Paul Riedel, a senior vice president at Starbucks, told Fortune in a statement. “This is the kind of technology that sparks creativity and helps customers discover something new.” 

Companies like Starbucks that are experimenting with AI recommendations are doing so as Americans start pushing back against interacting with the technology in all aspects of their lives. From resisting its use at work to rejecting AI-generated digital content, consumers are starting to feel AI fatigue, even anger. Brands are aware, with some recently pulling AI-created ads that received criticism online. A Harris Poll survey released this week found that only 39% of American consumers trust AI to make everyday purchases for them, with a large majority still skeptical of product quality when it is presented online.

Starbucks itself has also gone to great lengths to distance itself from the automation narrative in favor of preserving face-to-face interactions with customers. Last year, CEO Brian Niccol announced more locations would be increasing barista headcounts with fewer automated servings options, and called for more personal touches such as “handwritten notes” accompanying orders and using ceramic cups rather than disposable ones.

To be sure, Starbucks customers could surely do with some light guidance when the time comes to order. Between customizable sizes, number of espresso shots, sweetener types, and all the extra garnishes the coffee shop chain is known for, a simple latte order can sprawl into more than 300 billion distinct options. The enormous list of possible combinations has been cited as one reason behind long wait times during busy hours. After CEO Niccol took the job in 2024, one of his first moves was to slash menu size by 30%, calling many of the chain’s offerings “overly complex.”

AI recommendations might make it easier for customers to grapple with plentiful options, whether they are looking for a unique cut of sweater or a particular shade of caffeinated drink. Companies are trying to turn that decision fatigue into a competitive advantage, but if customers are pushed to order a specific blend of a caramel latte every time they get a coffee craving, they might miss the unexpected joys of a more free-flowing browse.