Starbucks Pilots ChatGPT-Powered Drink Recommendation Tool as AI Reshapes the Restaurant Industry

Starbucks is testing a new beta feature powered by OpenAI’s ChatGPT that recommends beverages to customers based on their mood, taste preferences, and personal goals — marking one of the most prominent integrations of generative artificial intelligence into a major fast-food ordering experience. While the company frames the tool as a way to simplify its famously sprawling menu, nutrition and data privacy experts are raising concerns about potential hidden downsides for consumers.

The pilot program represents a significant step in how one of the world’s largest restaurant chains intends to use generative AI at the point of sale. Users of the Starbucks mobile app — which already accounts for roughly 30 percent of all U.S. transactions at the company’s more than 16,000 domestic locations — can interact with the ChatGPT-powered assistant before placing their order. The conversational interface asks customers about their current emotional state, flavor preferences, dietary restrictions, and goals such as boosting energy or relaxing. Based on those inputs, the AI generates a personalized drink recommendation that can be ordered directly through the app. In a company with a menu so vast that baristas themselves often struggle to keep up with the permutations of syrups, milks, toppings, and sizes, the tool is positioned as a digital concierge of sorts. Starbucks has long invested in technology to streamline ordering; its Deep Brew AI platform, introduced several years ago, already powers personalized marketing and inventory management behind the scenes. The ChatGPT integration, however, is the company’s most consumer-facing AI deployment to date.

Yet the enthusiasm around AI-driven recommendations is not without scrutiny. Nutrition experts have flagged a fundamental concern: when an algorithm suggests a drink based on mood — say, recommending a Caramel Ribbon Crunch Frappuccino to someone who says they feel “stressed” or “need a treat” — there is a risk that emotional cues could be exploited to promote higher-sugar, higher-calorie options that also tend to carry higher price tags. Some of Starbucks’ most popular specialty drinks contain upward of 50 grams of sugar and exceed 400 calories, rivaling many fast-food meals. Without transparency about whether the recommendation engine factors in profit margins, promotional items, or inventory priorities alongside genuine customer wellness, critics argue consumers may be nudged in directions that serve the company’s bottom line more than their health. Data privacy advocates have also raised questions about how mood and behavioral data collected through these interactions will be stored, used, or potentially shared, particularly as AI systems become more sophisticated at building psychological profiles over time. Starbucks has not publicly detailed the data retention policies specific to this beta feature.

The broader implications of this pilot extend well beyond coffee. If successful, Starbucks’ integration could serve as a template for how major consumer brands deploy conversational AI to influence purchasing decisions at scale. The Federal Trade Commission has signaled growing interest in how AI-powered recommendation systems may constitute unfair or deceptive practices if they prioritize corporate revenue over consumer welfare without adequate disclosure. Meanwhile, the European Union’s AI Act, which began phased implementation in 2024, categorizes certain emotion-recognition technologies as high-risk, potentially complicating any international rollout of mood-based recommendation tools. For Starbucks, the stakes are considerable: the company’s Rewards loyalty program boasts more than 34 million active U.S. members, representing a massive dataset that, combined with AI-driven mood inputs, could create one of the most detailed consumer behavior profiles in the restaurant industry. Whether this pilot remains a novelty or scales into a core feature will likely depend on customer adoption rates, regulatory scrutiny, and whether the company can demonstrate that its AI genuinely serves the customer’s stated goals — not just the quarterly earnings report.